1.3 Geek Up—Tech Is Everywhere and You’ll Need It to Thrive
Learning Objectives
Appreciate the degree to which technology has permeated every management discipline.
See that tech careers are varied, richly rewarding, and poised for continued growth.
Shortly after the dot-com bubble, there was a lot of concern that tech jobs would be outsourced, leading many to conclude that tech skills carried less value and that workers with tech backgrounds had little to offer. Turns out this thinking was stunningly wrong. Tech jobs boomed, and as technology pervades all other management disciplines, tech skills are becoming more important, not less. Today, tech knowledge can be a key differentiator for the job seeker. It’s the worker without tech skills who needs to be concerned.
As we’ll present in depth in a future chapter, there’s a principle called Moore’s Law that’s behind fast/cheap computing. And as computing gets both faster and cheaper, it gets “baked into” all sorts of products and shows up everywhere: in your pocket, in your vacuum, and on the radio frequency identification (RFID) tags that track your luggage at the airport.
Well, there’s also a sort of Moore’s Law corollary that’s taking place with people, too. As technology becomes faster and cheaper, and developments like open source software, cloud computing, generative artificial intelligence, software as a service (SaaS), and outsourcing push technology costs even lower, tech skills are being embedded inside more and more job functions. And ubiquitous tech fuels our current era of “Big Data,” where bits-based insights move decision-making from hunch to science. What this means is that, even if you’re not expecting to become the next Tech Titan, your career will doubtless be shaped by the forces of technology. Make no mistake about it—there isn’t a single modern managerial discipline that isn’t being deeply and profoundly impacted by tech.
Finance
Many business school students who study finance aspire to careers in investment banking. Many i-bankers will work on IPOs (initial public stock offerings), in effect helping value companies the first time these firms wish to sell their stock on the public markets. IPO markets need new firms, and the tech industry is a fertile ground that continually sprouts new businesses like no other. Other i-bankers will be involved in valuing merger and acquisition (M&A) deals, and tech firms are active in this space, too. The technology sector has become a major driver of global M&A activity. Leading tech firms are flush with cash and constantly on the hunt for new firms to acquire. In just five years, Google has bought a whopping 103 firms, IBM has bought sixty-four, Microsoft has bought sixty-three, Cisco has bought fifty-seven, and Intel has bought forty-eight! Yahoo! bought thirty-seven companies in a year and a half. Trader beware—software might take your job. Researchers in the United States and China have shown an AI approach that beat major trading indices as well as traditional investment portfolio allocation strategies. No wonder so-called robo-advisors like Betterment have been growing by billions of dollars. And even in nontech industries, technology impacts nearly every endeavor as an opportunity catalyst or a disruptive wealth destroyer. The aspiring investment banker who doesn’t understand the role of technology in firms and industries can’t possibly provide an accurate guess at how much a company is worth.
Those in other finance careers will be lending to tech firms and evaluating the role of technology in firms in an investment portfolio. Most of you will want to consider tech’s role as part of your personal investments. And modern finance simply wouldn’t exist without tech. When someone arranges for a bridge to be built in Shanghai, those funds aren’t carried over in a suitcase—they’re digitally transferred from bank to bank. And forces of technology blasted open the two-hundred-year-old floor trading mechanism of the New York Stock Exchange, in effect forcing the NYSE to sell shares in itself to finance the acquisition of technology-based trading platforms that were threatening to replace it. Computer-automated trading, where a human doesn’t touch the deal at all, is responsible for some 60 percent of U.S. equity trading volume. For a look at the importance of tech in finance, consider that banking giant Chase spends $4 billion a year on technology and employs over twelve thousand technologists, globally. Tech isn’t a commodity for finance—it’s the discipline’s lifeblood.
How AI Is Powering the Future of Financial Services, JPMorgan Chase & Co.
Hear JPMorgan Chase's Sameena Shah discuss “How AI Is Powering the Future of Financial Services” as part of the MIT Technology Review EmTech Digital Conference.
Accounting
If you’re an accountant, your career is built on a foundation of technology. The numbers used by accountants are all recorded, stored, and reported by information systems, and the reliability of any audit is inherently tied to the reliability of the underlying technology. Increased regulation raises executive and board responsibility and ties criminal penalties to certain accounting and financial violations. Legislation has raised the stakes for mismanagement and misdeeds related to a firm’s accounting activities and have ratcheted up the importance of making sure accountants (and executives) get their numbers right. Negligence could mean jail time. This means the link between accounting and tech has never been tighter, and the stakes for ensuring systems accuracy have never been higher.
Business students might also consider that while accounting firms regularly rank near the top of Bloomberg Businessweek’s “Best Places to Start Your Career” list, many of the careers at these firms are highly tech-centric. Every major accounting firm has spawned a tech-focused consulting practice, and in many cases, these firms have grown to be larger than the accounting services functions from which they sprang. Today, Deloitte’s tech-centric consulting division is larger than the firm’s audit, tax, and risk practices. At the time of its spin-off, Accenture was larger than the accounting practice at former parent Arthur Andersen (Accenture executives are also grateful they split before Andersen’s collapse in the wake of the prior decade’s accounting scandals). Now, many accounting firms that had previously spun off technology practices are once again building up these functions, finding strong similarities between the skills of an auditor and skills needed in emerging disciplines such as information security and privacy.
Will AI Replace Accountants?
Interested in a career in accounting? Then you might be interested in this interview with Dr. Mike Willis, the head of the University of Cambridge (UK) Director of Accounting Grad Program, on the future of the accounting profession and the impact of AI on the field.
Marketing
Technology has thrown a grenade onto the marketing landscape, and as a result, the skill set needed by today’s marketers is radically different from what was leveraged by the prior generation. Online channels have provided a way to track and monitor consumer activities, and firms are leveraging this insight to understand how to get the right product to the right customer, through the right channel, with the right message, at the right price, at the right time. The success or failure of a campaign can often be assessed immediately based on online activity such as website visit patterns and whether a campaign results in an online purchase.
The ability to track customers, analyze campaign results, and modify tactics has amped up the return on investment of marketing dollars, with firms increasingly shifting spending from tough-to-track media such as print, radio, and television to the Web. And new channels continue to emerge: smartphone, tablet, smart TV, smartwatch and other wearables, smart auto, and more. Look to Apple to show how fast things grow: Within roughly four years, iOS devices were in the hands, backpacks, purses, and pockets of over two hundred million people worldwide, delivering location-based messages and services and even allowing for cashless payment. Apple has since sold over two billion iPhones worldwide, creating the primary channel for all sorts of customer engagement.
The rise of social media is also part of this blown-apart marketing landscape. Now all customers can leverage an enduring and permanent voice, capable of broadcasting word-of-mouth influence in ways that can benefit and harm a firm. Savvy firms are using social media to generate sales, improve their reputations, better serve customers, and innovate. Social media has also lead to the $21 billion+ influencer economy, while AI has ushered in the era of virtual influencers who have garnered millions of online followers and earned millions of promotional dollars along the way. Those who don’t understand this landscape risk being embarrassed, blindsided, and out of touch with their customers.
Search engine marketing (SEM), search engine optimization (SEO), customer relationship management (CRM), personalization systems, and a sensitivity to managing the delicate balance between gathering and leveraging data and respecting consumer privacy are all central components of the new marketing toolkit. And there’s no looking back—tech’s role in marketing will only grow in prominence. Analyst firm Gartner predicts that chief marketing officers are on a path to spend more on technology than any other function within the firm.
Computer-Generated Influencers on Social Media Give a Peek into the Future of Marketing
See how virtual influencers have garnered millions of followers. Also learn the term uncanny valley.
Transcript| 0.63 to 3.7199999999999998 | - That's Lil McKayla and Bermuda and Emma. |
| 3.7199999999999998 to 6.42 | Social media stars with millions of followers, |
| 6.42 to 9.48 | sharing selfies with friends, posting funny videos, |
| 9.48 to 14.07 | typical teen stuff, except these aren't teens. Hey wanna |
| 14.07 to 16.02 | - Be friends. Who does |
| 16.02 to 17.02 | - That? |
| 17.02 to 18.21 | They're not even human. |
| 18.21 to 22.23 | They're computer generated robots, virtual influencers |
| 22.23 to 23.82 | with plenty of fans. |
| 23.82 to 27.24 | - As soon as I clicked on her page, I think I was just |
| 27.24 to 28.83 | first obviously confused, |
| 28.83 to 31.5 | but then upon just looking at it, it was just, |
| 31.5 to 32.85 | it was just so cool to me. |
| 32.85 to 34.74 | - Carla Montgomery has been following little |
| 34.74 to 36.09 | McKayla for years. |
| 36.09 to 39.6 | - She like lives as a robot in our world, |
| 39.6 to 42.12 | and I think, you know, like her skincare routine, |
| 42.12 to 44.55 | she is like, well, sorry guys, it's just WD 40. |
| 44.55 to 46.32 | I think it's just super unique |
| 46.32 to 48.27 | - And super realistic. |
| 48.27 to 50.37 | - In looking at little Miguel, it's quite interesting. |
| 50.37 to 53.1 | - Carl McDormand studies human computer interaction |
| 53.1 to 56.13 | and explains the theory of the uncanny valley. |
| 56.13 to 58.26 | This graph explains what that is. |
| 58.26 to 60.9 | When a person looks at a series of virtual faces, |
| 60.9 to 62.105 | they feel comfortable. |
| 62.105 to 63.45 | As those faces become more |
| 63.45 to 66.36 | and more real, looking up to a certain point right |
| 66.36 to 68.61 | before they look a hundred percent human, |
| 68.61 to 71.13 | that's when people start feeling uncomfortable. |
| 71.13 to 73.98 | And it's that cute to creepy reaction |
| 73.98 to 76.47 | that researchers call the uncanny valley, |
| 76.47 to 78.09 | but something else is happening |
| 78.09 to 80.13 | with these virtual influencers. |
| 80.13 to 82.62 | - People get used to characters. |
| 82.62 to 86.01 | So if they're interacting with a particular |
| 86.01 to 90.66 | virtual influencer frequently, then |
| 90.66 to 92.91 | it may have been uncanny at some point, |
| 92.91 to 95.04 | but over time they'll get used to it. |
| 95.04 to 96.84 | - The interactions can be entertaining, |
| 96.84 to 98.88 | but researchers worry about the risk. |
| 98.88 to 103.88 | - I think there are general concerns about social media |
| 104.31 to 107.76 | kind of shaping young minds and young brains, |
| 107.76 to 110.76 | and I think the concerns would be even greater |
| 110.76 to 113.01 | with the digital influencers |
| 113.01 to 117.15 | because they're exploiting the human interface, which is |
| 117.15 to 119.61 | to say they look like people. |
| 119.61 to 122.67 | And so this gives them a lot of leverage. |
| 122.67 to 127.38 | It can be designed very specifically to be liked, |
| 127.38 to 130.11 | to be credible and so on. |
| 130.11 to 133.26 | And that's a kind of new danger that we haven't seen |
| 133.26 to 134.85 | before in society. |
| 134.85 to 137.28 | - These influencers creating relationships, |
| 137.28 to 140.91 | sharing stories appearing in the real world alongside real |
| 140.91 to 143.19 | people like Virtual Star Ray, |
| 143.19 to 145.59 | showing up on this magazine cover. |
| 145.59 to 149.73 | So if you look, can you tell who's real and who's digital? |
| 149.73 to 153.48 | - It questions the line of reality versus virtual reality. |
| 153.48 to 156.33 | - Virginia Tech professor Donna Alek thinks these virtual |
| 156.33 to 158.435 | influencers are becoming more popular |
| 158.435 to 161.79 | because of their targeted approach toward younger audiences, |
| 161.79 to 163.83 | which can really pay off. |
| 163.83 to 166.35 | Influencer marketing overall is a huge industry |
| 166.35 to 169.68 | with brands expected to spend up to $15 billion on it |
| 169.68 to 172.59 | by next year, and these digital stars are coming |
| 172.59 to 174.48 | for a bigger piece of that pie, |
| 174.48 to 177.36 | making money from sponsored posts, concerts, |
| 177.36 to 179.44 | even virtual reality video games. |
| 179.44 to 182.95 | Lil McKayla reportedly made more than $11 million last year. |
| 182.95 to 185.92 | - It's translated now to somebody they can relate to. |
| 185.92 to 188.17 | Someone that looks like them, talks like them, |
| 188.17 to 190.45 | feels like them, and then them translating |
| 190.45 to 193.27 | to be an influencer for these young communities. |
| 193.27 to 196.605 | - There's a company behind that looking |
| 196.605 to 198.195 | to sell something, right? |
| 198.195 to 200.32 | Like this comes back to like branding |
| 200.32 to 202.27 | and influencing at its very core. |
| 202.27 to 205.99 | - Absolutely. And it is the wizard behind the curtain |
| 205.99 to 208.84 | that is playing like a puppeteer |
| 208.84 to 211.24 | with these virtual beings. |
| 211.24 to 212.38 | You're not only getting into it, |
| 212.38 to 213.55 | you're getting into you know |
| 213.55 to 214.99 | who they're friends with, who they're not. |
| 214.99 to 216.4 | And then by the way, I'm using this |
| 216.4 to 217.69 | product and that product. |
| 217.69 to 220.45 | - One of the companies creating virtual influencers |
| 220.45 to 223.15 | emphasized it's a new way to story tell, |
| 223.15 to 226.15 | but did not respond to a longer list of questions from us. |
| 226.15 to 229.33 | Another gave us a lengthy statement from its CEO saying, |
| 229.33 to 231.82 | in part, the next generation of young people |
| 231.82 to 234.16 | who can smell whether something seems authentic |
| 234.16 to 236.29 | or not, will be the main players in the |
| 236.29 to 237.58 | age of the metaverse. |
| 237.58 to 241.63 | Adding virtual humans are a mirror to humans to teach us |
| 241.63 to 243.34 | what we really should treasure. |
| 243.34 to 245.95 | What do parents need to know about the rise |
| 245.95 to 247.845 | of virtual influencers online? Talk |
| 247.845 to 250.18 | - To your kids and find out what's going on. |
| 250.18 to 252.85 | Get involved with them. We need to be informed as parents |
| 252.85 to 254.68 | and go a little bit beyond just the birds |
| 254.68 to 258.01 | and bees these days because there's a lot more influencers |
| 258.01 to 260.8 | and impact out there that could really change the future |
| 260.8 to 262.93 | of our children based on how they're being conditioned. |
| 265.36 to 267.34 | - Thanks for watching our YouTube channel. |
| 267.34 to 268.875 | Follow today's top stories |
| 268.875 to 272.29 | and Breaking news by downloading the NBC News app. |
Operations
A firm’s operations management function is focused on producing goods and services, and operations students usually get the point that tech is the key to their future. Quality programs, process redesign, supply chain management, factory automation, and service operations are all tech-centric. These points are underscored in this book as we introduce several examples of how firms have designed fundamentally different ways of conducting business (and even entirely different industries), where value and competitive advantage are created through technology-enabled operations.
Google Cloud Manufacturing Operations Solutions
This short video from Google Cloud platform shows how Big Data, robotics, and other technologies are transforming the factory floor, and how goods are produced for market.
Human Resources
Technology helps firms harness the untapped power of employees. Knowledge management systems are morphing into social media technologies—social networks, wikis, and Twitter-style messaging systems that can accelerate the ability of a firm to quickly organize and leverage teams of experts. And crowdsourcing tools and question-and-answer sites like Quora and Stack Overflow allow firms to reach out for expertise beyond their organizations. Human resources (HR) directors are using technology for employee training, screening, and evaluation. The accessibility of end-user technology means that every employee can reach the public, creating an imperative for firms to set policy on issues such as firm representation and disclosure and to continually monitor and enforce policies as well as capture and push out best practices. The successful HR manager recognizes that technology continually changes an organization’s required skill sets as well as employee expectations.
The hiring and retention practices of the prior generation are also in flux. Recruiting hasn’t just moved online; it’s now grounded in information systems that scour databases for specific skill sets, allowing recruiters to cast a wider talent net than ever before. Job seekers are writing résumés with keywords in mind, aware that the first cut is likely made by a database search program, not a human being. The rise of professional social networks also puts added pressure on employee satisfaction and retention. Prior HR managers fiercely guarded employee directories for fear that a headhunter or competitive firm might raid top talent. Now the equivalent of a corporate directory can be easily pulled up via LinkedIn, a service complete with discrete messaging capabilities that can allow competitors to rifle-scope target your firm’s best and brightest. Thanks to technology, the firm that can’t keep employees happy, engaged, and feeling valued has never been more vulnerable.
And while many students have been wisely warned that inappropriate social posts can ruin their job candidacy, also know that the inverse is true. In many ways social media is “the new résumé.” Thoughtful blog posts, a compelling LinkedIn presence, Twitter activity reflecting an enthusiastic and engaged mind, and, for tech students, participation in collaborative coding communities like GitHub all work to set apart a candidate from the herd. If you can’t be found online, some employers may wonder if you have current skills, or if you have something to hide.
How Companies, Human Resource Departments Are Using AI When Hiring, Plus a Look at Regulation
Yahoo! Finance discusses new laws and a shifting set of concerns for organizations that are using more technology as part of their human resources functions and especially their hiring processes.
Transcript| 2.61 to 4.02 | - It is well known |
| 4.02 to 7.86 | that a human is not always the first one you see when |
| 7.86 to 9.815 | to resume your apply to a new job. |
| 9.815 to 13.2 | Often it's a computerized system that scans your resume |
| 13.2 to 15.42 | for the relevant skills and qualifications. |
| 15.42 to 16.44 | I don't know if everybody knows |
| 16.44 to 20.46 | that nearly one in four organizations use AI |
| 20.46 to 23.91 | to support HR activities according to the Society |
| 23.91 to 26.1 | for Human Resource Management. |
| 26.1 to 30.69 | Of those nearly eight in 10 use AI tools for hiring. |
| 30.69 to 33.9 | Now those organizations are getting hit with some new laws. |
| 33.9 to 36.33 | They must comply with Yahoo Finances. |
| 36.33 to 38.82 | Carry Hannin has that story. Hi there, Carrie. |
| 38.82 to 40.5 | What are we, what are we hearing? |
| 40.5 to 42.81 | - Hey, Dave. Yeah, well, you know, it's been a bit |
| 42.81 to 46.145 | of the Wild West out there in terms of regulations |
| 46.145 to 49.5 | and restrictions on these AI tools |
| 49.5 to 50.91 | that employers use in hiring. |
| 50.91 to 52.65 | I mean, yeah, there's things like, you know, |
| 52.65 to 54.93 | they can game the system for your, you know, trying |
| 54.93 to 56.19 | to get the right keywords, |
| 56.19 to 58.32 | but there's some inherent bias |
| 58.32 to 61.59 | that can built be built into these for, you know, race |
| 61.59 to 63.66 | and gender and all kinds of things. |
| 63.66 to 65.61 | So, you know, we haven't had seen a lot |
| 65.61 to 67.29 | of regulation out there to date, |
| 67.29 to 71.76 | but in our, in April in New York City for the first time, |
| 71.76 to 73.38 | there's going to be a law |
| 73.38 to 75.75 | that actually addresses this issue. |
| 75.75 to 77.37 | And so what's gonna happen there, |
| 77.37 to 79.11 | and it sort of generally looks at |
| 79.11 to 82.05 | how most employers are using these AI tools |
| 82.05 to 85.98 | and it's gonna really put it on the employer |
| 85.98 to 88.955 | to be transparent, to tell people, Hey, this is |
| 88.955 to 92.58 | what we're looking for, the qualifications, these, you know, |
| 92.58 to 94.44 | AI tools are gonna be scanning for. |
| 94.44 to 96.24 | These are the characteristics, |
| 96.24 to 98.43 | the job qual qualifications you need. |
| 98.43 to 101.04 | And if someone feels that it's not going to, |
| 101.04 to 104.73 | they're not able to be judged fairly in the system, |
| 104.73 to 107.52 | they have the right to ask to try a different way. |
| 107.52 to 108.93 | Now that gets a little tricky. |
| 108.93 to 110.82 | The whole thing is rather complicated |
| 110.82 to 114.3 | and it's going to be pretty messy as we move forward. |
| 114.3 to 116.19 | But, so New York City, that's the first one. |
| 116.19 to 118.08 | It was supposed to start in January, |
| 118.08 to 121.08 | but employers are scrambling to figure out what do they need |
| 121.08 to 122.555 | to do to find out |
| 122.555 to 124.95 | and do these audits on their systems |
| 124.95 to 127.86 | to see if there is bias that's happening. |
| 127.86 to 129.18 | And we all can complain about it, |
| 129.18 to 130.74 | but they are gonna be the ones |
| 130.74 to 132.69 | that will be responsible if there are |
| 132.69 to 134.64 | lawsuits on discrimination. |
| 134.64 to 135.935 | - Certainly some changes coming there. |
| 135.935 to 138.81 | Carrie, what are we hearing just about the rise in AI aided |
| 138.81 to 140.225 | video interviews? |
| 140.225 to 142.5 | There's a little bit more talk about that. |
| 142.5 to 144.78 | What can you tell us there? And just there's some potential |
| 144.78 to 146.94 | shortfalls you would also think We'll see. |
| 146.94 to 148.95 | - Exactly. I'm so glad you asked that. |
| 148.95 to 151.86 | Now this is something that the EEOC, you know, |
| 151.86 to 156.3 | the US Equal Employment Opportunity Commission put out some |
| 156.3 to 159.15 | things, some guidance on this, particularly for people |
| 159.15 to 161.375 | with disabilities because, oh my gosh, |
| 161.375 to 164.7 | this has become the rage now that you're interviewed by sort |
| 164.7 to 167.43 | of a chat bot on this automated, |
| 167.43 to 168.9 | you don't even have a person. |
| 168.9 to 171.84 | There's no sort of, it's really a bizarre experience. |
| 171.84 to 176.16 | But what it does is people are judged on their speech |
| 176.16 to 179.74 | patterns, the words they use, their facial muscles, |
| 179.74 to 182.86 | all kinds of things that the AI can, you know, |
| 182.86 to 184.57 | the tool can judge you on. |
| 184.57 to 188.2 | Now, if you have a disability, say for example, |
| 188.2 to 191.44 | a speech impediment, that can really come up |
| 191.44 to 194.17 | to really haunt you in these sort of, in interviews |
| 194.17 to 195.19 | and you can be discarded |
| 195.19 to 198.28 | because they use the way your, the rhythm of your speech, |
| 198.28 to 203.2 | the cadence to determine if you can make good decisions now. |
| 203.2 to 204.76 | So that is something that, you know, |
| 204.76 to 206.65 | opens up a whole other can of worms |
| 206.65 to 209.89 | and I think we're gonna see a lot more moving forward in |
| 209.89 to 211.96 | that area as well. |
| 211.96 to 213.91 | - All right. Great stuff Kerry Hannan, thanks so much. |
The Law
And for those looking for careers in corporate law, many of the hottest areas involve technology. Intellectual property, patents, piracy, and privacy are all areas where activity has escalated dramatically in recent years. The number of U.S. patent applications awaiting approval has tripled in the past decade, while China saw a threefold increase in patent applications in just five years. Firms planning to leverage new inventions and business methods need legal teams with the skills to sleuth out whether a firm can legally do what it plans to. Others will need legal expertise to help them protect proprietary methods and content, as well as to help enforce claims in the home country and abroad. Tech is increasingly being used to automate rote legal tasks, and generative AI is becoming mainstream in the legal profession; however, we’ve also seen examples of clueless lawyers who have violated professional practice standards by submitting ChatGPT-generated legal analyses that presented bogus, non-existent examples.
How CoCounsel's AI Assistant Is Making Lawyers Unstoppable!
Watch a brief demonstration of CoCounsel, one of many generative AI tools that are enhancing how attorneys do their jobs.
Transcript| 0 to 2.67 | - The best way to demonstrate that is to jump right in |
| 2.67 to 4.44 | and take a look at co-counsel. |
| 4.44 to 5.64 | So let's jump right in. |
| 5.64 to 8.01 | - This is co-counsel. I'll start here like |
| 8.01 to 9.45 | with other chat applications |
| 9.45 to 10.95 | that people may have been aware of. |
| 10.95 to 12.515 | If you want kind of ask coun |
| 12.515 to 13.83 | to do anything for you via chat. |
| 13.83 to 18.39 | So you can say, write up a quick summary why legal tech |
| 18.39 to 22.56 | enthusiasts should be excited about large language models. |
| 22.56 to 24.72 | We've programmed co-counsel to have a number |
| 24.72 to 25.86 | of legal specific skills. |
| 25.86 to 27.96 | The six that we're launching with are things like asking |
| 27.96 to 30.21 | to review documents, to do legal research, |
| 30.21 to 33.12 | to extract data from a contract to summarize large |
| 33.12 to 35.64 | and lengthy difficult documents, other similar skills. |
| 35.64 to 37.77 | So to give you a sense of how this works in practice, |
| 37.77 to 40.71 | I might ask my AI to do legal research for me |
| 40.71 to 44.67 | and say, what is the pleading standard allegations under the |
| 44.67 to 47.16 | False Claims Act in the ninth circuit? |
| 47.16 to 49.14 | I'm just talking to this like I would a colleague. |
| 49.14 to 52.08 | What's happening behind the scenes is it's deciding |
| 52.08 to 54.12 | what search queries to run based on this query. |
| 54.12 to 56.31 | It is running those search queries against our database |
| 56.31 to 59.04 | of daily updating cases, statutes, rules and regulations. |
| 59.04 to 60.72 | It is reading hundreds of results |
| 60.72 to 61.98 | that come back from these searches |
| 61.98 to 63.03 | and then it's gonna use that |
| 63.03 to 66.09 | to put together a legal research memo in a response. |
| 66.09 to 69.9 | - How can lawyers who are using this know that the result |
| 69.9 to 71.64 | that comes back is reliable? |
| 71.64 to 74.49 | - It's a great question. The short answer is two pieces. |
| 74.49 to 77.58 | The first is I'll kick off a doc review project |
| 77.58 to 82.14 | that's gonna kind of similar, uploading 33 speeches |
| 82.14 to 84.15 | that Barack Obama's given |
| 84.15 to 85.59 | and you can ask any questions in it. |
| 85.59 to 89.88 | Does Obama make a joke? Does Obama mention his daughter's? |
| 89.88 to 91.29 | - Does he mention his dog? |
| 91.29 to 94.59 | - Does the speech include veiled |
| 94.59 to 96.39 | criticism of Republicans? |
| 96.39 to 98.37 | What this is gonna do is bring back |
| 98.37 to 99.42 | answers relatively quickly. |
| 99.42 to 101.79 | It's also gonna empower you to find the exact page |
| 101.79 to 103.83 | where it claims it got the answer from for you to verify |
| 103.83 to 105.27 | for yourself for all of the skills. |
| 105.27 to 107.555 | It's going to answer questions based on the |
| 107.555 to 108.84 | cases that it's itself reading. |
| 108.84 to 111.78 | It's also going to answer questions based on the documents |
| 111.78 to 114.15 | that it's reading so you can verify that entirely yourself. |
| 114.15 to 115.38 | There's also another set of things |
| 115.38 to 117.63 | that we do on the product side to make sure |
| 117.63 to 120.72 | that it's answering based solely on the information |
| 120.72 to 121.92 | that it's, that it's reading, |
| 121.92 to 124.65 | that it's not pulling from extraneous information from kind |
| 124.65 to 126.93 | of from its memory already it's, it's pulled dozens |
| 126.93 to 129.12 | of cases will in a moment kind |
| 129.12 to 131.94 | of start writing it's reason why it believes this case is |
| 131.94 to 133.38 | relevant for every single of these cases. |
| 133.38 to 135.605 | - In addition to actually generating the memo, |
| 135.605 to 139.44 | it's also showing you the cases from which it has drawn the |
| 139.44 to 140.58 | expertise in the memo |
| 140.58 to 142.8 | and the sections of those cases that are relevant. |
| 142.8 to 145.38 | - That's exactly right. You can see right now is currently |
| 145.38 to 147.48 | typing out it's memorandum. |
| 147.48 to 148.62 | This is the kind of work that |
| 148.62 to 150.24 | as an associate would take me hours. |
| 150.24 to 151.35 | It's giving me this deep |
| 151.35 to 152.73 | and nuanced nuance based on the career. |
| 152.73 to 153.84 | - I think some people worry |
| 153.84 to 155.315 | that this would replace a lawyer, |
| 155.315 to 158.52 | but really what it's doing is giving you an an extremely |
| 158.52 to 162.065 | good place to start with the memo as well as the cases so |
| 162.065 to 165.15 | that you can go and look at them and tweak the memo |
| 165.15 to 169.08 | and save many hours of time that you would've spent. |
| 169.08 to 171.9 | It still requires the human to look over it in the end |
| 171.9 to 173.765 | and ensure that it's expressed the way |
| 173.765 to 175.23 | that it would like to be expressed. |
| 175.23 to 176.88 | - You go a step further than that. First of all, I think |
| 176.88 to 178.71 | that lawyer is empowered with this are gonna be able to do |
| 178.71 to 180.04 | so much more, so much better |
| 180.04 to 181.9 | and so much faster for their clients |
| 181.9 to 183.91 | that they're gonna be in much more demand |
| 183.91 to 185.41 | and gonna be put in a position |
| 185.41 to 187.27 | to do really great work for their clients. |
| 187.27 to 188.65 | I think that's gonna be really empowering. |
| 188.65 to 189.94 | They're not gonna see less work |
| 189.94 to 191.26 | and see a lot more demand for their work |
| 191.26 to 192.37 | at a cheaper lower price. |
| 192.37 to 194.32 | I also think it's actually knowing what questions to ask |
| 194.32 to 196.72 | and knowing what to look for in the documents, what to look |
| 196.72 to 198.88 | for in the cases, how to conduct that legal research. |
| 198.88 to 201.13 | I think the new skill is going to be about deciding |
| 201.13 to 203.83 | how do you delegate to this incredibly powerful asset |
| 203.83 to 204.94 | that is going through hundreds |
| 204.94 to 207.16 | or thousands of pages of text incredibly quickly. |
| 207.16 to 208.24 | So for example, for every one |
| 208.24 to 210.34 | of these it's saying did Obama make a joke? |
| 210.34 to 211.75 | He jokes about not having remarks in front of him |
| 211.75 to 213.82 | and also has a letter to to talk about this presidency. |
| 213.82 to 216.37 | - But additional information is so critical for lawyers |
| 216.37 to 218.47 | and so useful to get that on the page. |
| 218.47 to 221.2 | - Exactly. It's kind of explainability, including pointing |
| 221.2 to 222.67 | to the exact right page in the case |
| 222.67 to 226.03 | and even open up the page four of this speech to confirm |
| 226.03 to 227.86 | or deny is a really critical aspect |
| 227.86 to 229.12 | of working with this kind of ai. I |
| 229.12 to 231.58 | - Can see that in some cases it said insufficient. |
| 231.58 to 232.87 | So if you ask it a question |
| 232.87 to 235.57 | and it does not know the answer, will it actually say, |
| 235.57 to 236.59 | I don't know the answer |
| 236.59 to 238.72 | or I have insufficient information in order |
| 238.72 to 240.075 | to respond to this question. |
| 240.075 to 241.57 | - That's exactly right. That's exactly what it's doing. |
| 241.57 to 244.96 | - I had heard that there was a limitation to the number |
| 244.96 to 248.895 | of pages that this type of technology is able to read, |
| 248.895 to 251.89 | but you've just added many pages worth of speeches. |
| 251.89 to 254.14 | Have you architected something on the backend |
| 254.14 to 257.8 | that allows co-counsel to read many pages |
| 257.8 to 259.15 | of documents? Is there a limit? |
| 259.15 to 261.19 | - Yeah, that's one of the major engineering breakthroughs |
| 261.19 to 263.83 | that we are working on to make this kind of technology work, |
| 263.83 to 267.135 | which is putting it in a position to read through thousands |
| 267.135 to 269.05 | of pages, a hundred thousand pages for the large ones |
| 269.05 to 271.93 | that work with us, they can read up to 500,000 pages |
| 271.93 to 274.12 | of documents overnight using this technology, |
| 274.12 to 276.01 | putting them in position to do an intense amount |
| 276.01 to 279.64 | of e-discovery or categorizing documents their CAM database. |
| 279.64 to 281.14 | So that's major breakthrough that as you're working |
| 281.14 to 282.67 | - On Jake, this is fascinating. |
| 282.67 to 284.895 | I noticed that one of the skills is summarization, |
| 284.895 to 287.62 | which is another use case we are seeing in the industry. |
| 287.62 to 291.28 | I'm very interested in the extracting data from a contract |
| 291.28 to 292.28 | - Here. |
| 292.28 to 293.5 | For example, I'm uploading, you know, |
| 293.5 to 294.855 | a few different commitment letters |
| 294.855 to 297.4 | and private equity deals for leveraged buyouts for loans. |
| 297.4 to 298.75 | You can ask any question like |
| 298.75 to 300.975 | who are the parties to the contract? |
| 300.975 to 304.15 | What is the termination fee? I can specify that. |
| 304.15 to 306.13 | I want that in the form of a number by the way, if it's, |
| 306.13 to 307.99 | it's an interpretation wrong to introducing kind |
| 307.99 to 309.46 | of everywhere, give you the opportunity |
| 309.46 to 310.63 | to re-edit your question |
| 310.63 to 312.19 | and that way you'll have a higher likelihood |
| 312.19 to 314.17 | of it really finding things you're looking for |
| 314.17 to 317.075 | - Helping you engineer prompts that are effective. Exactly. |
| 317.075 to 319.72 | - Very clever This, this is ultimately extremely flexible. |
| 319.72 to 321.045 | So you may have as a firm |
| 321.045 to 322.63 | or as an attorney your own set |
| 322.63 to 324.76 | of questions you wanna ask against contracts |
| 324.76 to 326.68 | and this will like document reviews, |
| 326.68 to 328.66 | pull out deep answers to those questions. |
| 328.66 to 330.855 | - I'm curious about a couple of other things. |
| 330.855 to 332.26 | One is security. |
| 332.26 to 334.605 | It's one of the things that law firms have been concerned |
| 334.605 to 337.75 | about with chat GBT, which is an open source. |
| 337.75 to 340.42 | Firms are often very concerned about security |
| 340.42 to 341.59 | around their own work product |
| 341.59 to 343.45 | and contracts within a transaction. |
| 343.45 to 347.05 | How can they feel safe using co-counsel on their own |
| 347.05 to 348.28 | internal work product? |
| 348.28 to 350.2 | - Being in legal tech for the last 10 years |
| 350.2 to 351.37 | and working with 60 |
| 351.37 to 352.84 | of the largest law firms in the trade means |
| 352.84 to 354.7 | that we totally get the security concerns |
| 354.7 to 356.29 | that firms may have around these kinds of things. |
| 356.29 to 358.015 | And part what's important |
| 358.015 to 359.51 | to distinguish is when you're working |
| 359.51 to 361.4 | with a public tool like A GPT, |
| 361.4 to 362.87 | every time you put information into it in |
| 362.87 to 364.73 | that model is learning the information |
| 364.73 to 365.78 | from your discussions with it. |
| 365.78 to 366.98 | Large companies like Amazon |
| 366.98 to 369.32 | and even Microsoft, which is a big investor in open ai, |
| 369.32 to 370.88 | the folks who make chat GPT, |
| 370.88 to 372.29 | so their employees don't put any |
| 372.29 to 373.52 | confidential information in here. |
| 373.52 to 375.11 | It can leak into the big model for us. |
| 375.11 to 378.59 | We have our own private servers that host these models. |
| 378.59 to 380.72 | It never learns by design. |
| 380.72 to 383.03 | Any of the information you're putting into it never retains |
| 383.03 to 384.29 | the information that you're putting into it. |
| 384.29 to 387.62 | The only way that the data comes in contact with our AI is |
| 387.62 to 388.7 | for the processing, for reading |
| 388.7 to 389.69 | and understanding of putting it |
| 389.69 to 390.8 | out and then it's gone forever. |
| 390.8 to 393.14 | That's a really, really important part of our architecture. |
| 393.14 to 395.24 | Frankly, firms should ask these kinds of questions. |
| 395.24 to 397.52 | There'll be many more in the future that attempt |
| 397.52 to 399.62 | to do these things using this brand of technology. |
| 399.62 to 402.08 | There is a major difference between the companies |
| 402.08 to 403.94 | that can provide that kind of level of security |
| 403.94 to 406.435 | and capabilities and those who cannot, firms have |
| 406.435 to 407.78 | to look into that when they're trying |
| 407.78 to 409.79 | to decide whether they leverage this really impressive |
| 409.79 to 411.62 | and exciting technology, which they should. |
| 411.62 to 412.82 | They have to look into whether |
| 412.82 to 414.38 | or not the model is being trained on the |
| 414.38 to 415.585 | documents we we put into it. |
| 415.585 to 418.13 | - Are you able to save a model effectively? |
| 418.13 to 419.42 | So a list of questions so |
| 419.42 to 421.34 | that you could reuse that for later. |
| 421.34 to 424.975 | And with data that is extracted, are you able to tell it |
| 424.975 to 428.6 | to populate a summary table so that it extracts data |
| 428.6 to 431.45 | and then also populates a table |
| 431.45 to 434.81 | or a database with the data that has been extracted? |
| 434.81 to 436.34 | - Absolutely. So what we do |
| 436.34 to 439.52 | for example here is I've uploaded these documents earlier. |
| 439.52 to 441.355 | I'm gonna select them and I can run the same |
| 441.355 to 442.52 | analysis of those documents. |
| 442.52 to 444.38 | It also saves previously used questions. |
| 444.38 to 446 | Those are the questions that we asked earlier. |
| 446 to 448.22 | You can heart them so that you at the top always I can kick |
| 448.22 to 450.65 | off a doc review with exactly those kinds of questions again |
| 450.65 to 451.88 | and we're gonna make it even easier in your |
| 451.88 to 453.17 | future to save templates. |
| 453.17 to 454.82 | In terms of getting the answers out, |
| 454.82 to 456.29 | we always allow you to download things. |
| 456.29 to 457.64 | For example, doing Excel spreadsheet |
| 457.64 to 460.31 | or in the case of like a legal research memo to download it |
| 460.31 to 462.68 | as a Word document and to you'll get, you know, |
| 462.68 to 465.38 | all this context, all the cases, the answer links |
| 465.38 to 466.79 | to the cases, the quotes in the cases |
| 466.79 to 467.87 | that are relevant, et cetera. |
| 467.87 to 470.33 | All in a a version that is easy for you to use. |
| 470.33 to 471.47 | - Oh, this is fantastic. Jake, |
| 471.47 to 474.38 | you're already working in beta with a number of large firms |
| 474.38 to 475.94 | and other entities as well. |
| 475.94 to 477.895 | - That's right. We found it very important over the last |
| 477.895 to 480.205 | nearly six months now to engage with folks |
| 480.205 to 483.44 | who are at large law firms, small law firms, nonprofits, |
| 483.44 to 485.99 | inhouse council, to really through its paces |
| 485.99 to 487.31 | and make sure that works really |
| 487.31 to 488.57 | well in the early days of the beta. |
| 488.57 to 491.48 | They, they caught issues with the way you interact with it, |
| 491.48 to 493.16 | what kind of skills we should really prioritize |
| 493.16 to 495.685 | and use, help it make it faster, all those kinds of things. |
| 495.685 to 497.48 | And then beta customers did a really great work |
| 497.48 to 499.1 | and had really great feedback for us to get |
| 499.1 to 500.15 | to the place we are now. |
| 500.15 to 501.8 | It puts us in a position that when we rolled out |
| 501.8 to 503.21 | to the broader market, a lot |
| 503.21 to 504.47 | of the kinks have been ironed out. |
| 504.47 to 506.395 | Still new technology still needs to be worked on |
| 506.395 to 508.85 | and always will be like we're constantly improving, |
| 508.85 to 510.5 | but a lot of the kings have been ironed out |
| 510.5 to 512.6 | and we get to come out with a pretty fully baked |
| 512.6 to 513.6 | - Product. |
| 513.6 to 515.9 | Well this is fantastic. It looks like co-counsel is the |
| 515.9 to 517.135 | lawyer's new best friend. |
| 517.135 to 518.66 | - Coco their best friend. Yeah, Coco, |
| 519.5 to 520.5 | - I love it. |
| 520.5 to 521.99 | Exactly like their mascot. Their mascot. Yeah. |
| 521.99 to 524.87 | As of the day this goes live March 1st, is it available? |
| 524.87 to 528.59 | Can firms access this call? You demo pilot? |
| 528.59 to 530.48 | - We're gonna be working with a smaller handful |
| 530.48 to 531.86 | of firms at the very beginning |
| 531.86 to 533.725 | because we wanna make sure that |
| 533.725 to 534.92 | as we deploy this technology, |
| 534.92 to 537.48 | it's un responsibly in a well-trained kind of manner |
| 537.48 to 540 | that we have the server capacity, which is very expensive |
| 540 to 541.5 | and difficult to obtain, to make sure |
| 541.5 to 543.33 | that everybody's using it gets a great experience. |
| 543.33 to 545.765 | But we'll be rolling it out to as many people as we can, |
| 545.765 to 547.59 | as fast as we can in that, making sure |
| 547.59 to 548.67 | that the first customers |
| 548.67 to 550.145 | who are onboarded have a great experience. |
| 550.145 to 552.03 | So if you're interested, come March 1st, |
| 552.03 to 553.92 | we'll have very easy ways to get in touch |
| 553.92 to 556.59 | and know that we're working around the clock to make it so |
| 556.59 to 558 | that after that first batch |
| 558 to 559.56 | or two of clients who come in early |
| 559.56 to 560.88 | and get rolled out, that we'll get |
| 560.88 to 562.74 | to the wider market as soon as we can. |
| 562.74 to 565.26 | - Jake, thank you so much for showing us co-counsel |
| 565.26 to 568.415 | or Coco, it looks like a groundbreaking new product, |
| 568.415 to 569.58 | so congratulations. |
| 569.58 to 570.21 | Thank you so much. |
Information Systems Careers
While the job market goes through ebbs and flows, recent surveys have shown there is no end in sight for workers with technology skills. In the U.S. News ranking of “Best Jobs in America” tech skills made up all of the top fifteen jobs that weren’t part of the health care industry. Glassdoor listed tech jobs in nine of the top ten slots in its “50 Best Jobs in America.” The World Economic Forum’s Future of Jobs Report ranked IT job titles in twelve of its top twenty, and demand for computer and information technology workers will outstrip supply for years to come. Information- and computer-related careers are expected to grow by double digits for at least the next decade. Want to work for a particular company? Chances are they’re looking for tech talent. Eighty-two percent of CEOs expect to need more tech workers. LinkedIn says, “In the next five years, we’ll see 150 million more technology-related jobs across industries globally, so demand for digital skills is very much on the rise.” The Harvard Business Review has declared “Data Scientist” the “Sexiest Job of the 21st Century,” and Bloomberg says undergraduates with data science training are now being hired with salaries greater than investment banking analysts. The recent pandemic not only showed the resilience of tech careers, but also the work-from-home flexibility offered to tech staff, who often just need a laptop and reliable Internet to get the job done. Fortune’s rankings of the “Best Companies to Work For” is full of technology firms and has been topped by tech businesses nearly every year since it started. And everyone wants to hire more coders from underrepresented groups. Apple, Etsy, Square, Meta, and Alphabet are among the firms with programs to prep and encourage more women and minorities to pursue tech careers (details in endnotes).
Students studying technology can leverage skills in ways that range from the highly technical to those that emphasize a tech-centric use of other skills. And why be restricted to just the classes taught on campus? Resources like Coursera, Codecademy, Udemy, edX, YouTube, and others provide a smorgasbord of learning where the smart and motivated can geek up. Carve out some time to give programming a shot—remember, the founders of Tumblr and Instagram were largely self-taught. Your author even offers hundreds of free videos on learning to code iOS apps from scratch, as well as tutorials on building robotics and other hardware products. The high demand for scarce technical talent has also led many tech firms to offer six-figure starting salaries to graduating seniors from top universities. Take some advice from the Harvard Business Review: “Leading a digital transformation? Learn to code.” Opportunities for programmers abound, particularly for those versed in new technologies. But there are also non-programming roles for experts in areas such as user-interface design (who work to make sure systems are easy to use), process design (who leverage technology to make firms more efficient), and strategy (who specialize in technology for competitive advantage). Nearly every large organization has its own information systems department. That group not only ensures that systems get built and keep running but also increasingly takes on strategic roles targeted at proposing solutions for how technology can give the firm a competitive edge. Career paths allow for developing expertise in a particular technology (e.g., business intelligence analyst, database administrator, social media manager), while project management careers leverage skills in taking projects from idea through deployment.
Even in consulting firms, careers range from hard-core programmers who “build stuff” to analysts who do no programming but might work identifying problems and developing a solutions blueprint that is then turned over to another team to code. Careers at tech giants like Apple, Google, and Microsoft don’t all involve coding end-user programs either. Each of these firms has its own client-facing staff that works with customers and partners to implement solutions. Field engineers at these firms may work as part of (often very lucratively compensated) sales teams to show how a given company’s software and services can be used. These engineers often put together prototypes that are then turned over to a client’s in-house staff for further development. An Apple field engineer might show how a firm can leverage iPads in its organization, while a Google field engineer can help a firm incorporate search, banner, and video ads into its online efforts. Careers that involve consulting and field engineering are often particularly attractive for those who are effective communicators who enjoy working with an ever-changing list of clients and problems across various industries and in many different geographies.
Upper-level career opportunities are also increasingly diverse. Consultants can become partners who work with the most senior executives of client firms, helping identify opportunities for those organizations to become more effective. Within a firm, technology specialists can rise to be chief information officer or chief technology officer—positions focused on overseeing a firm’s information systems development and deployment. And many firms are developing so-called C-level specialties in emerging areas with a technology focus, such as chief information security officer (CISO), and chief privacy officer (CPO). Senior technology positions may also be a ticket to the chief executive’s suite. Fortune pointed out how the prominence of technology provides a training ground for executives to learn the breadth and depth of a firm’s operations, and an understanding of the ways in which a firm is vulnerable to attack and where it can leverage opportunities for growth.
Generative AI and Programming, Peter Norvig, Director of Research, Google
No, tech jobs aren’t going away; tech will be baked into even more jobs. Peter Norvig, one of the world’s leading AI scientists, discusses the role of AI in the future of tech jobs, and how most jobs will become tech-enabled.
Transcript| 5.105 to 8.25 | - So my intro is I was doing AI before it was cool, |
| 10.11 to 12.99 | but I'm still here and, |
| 12.99 to 15.06 | and I'm gonna talk about programming today. |
| 15.06 to 18.065 | And you know, I've had a lot of roles in my career. |
| 18.065 to 23.015 | I've been a researcher, an author, a manager, a teacher. |
| 23.015 to 25.59 | But at my at heart, I'm really a programmer. |
| 25.59 to 27.12 | That's what I love to do. |
| 27.12 to 30.24 | So I wanna give you, I only got 15 minutes |
| 30.24 to 34.17 | that's less than dj and I'm not complaining, |
| 35.52 to 37.8 | but I still wanna squeeze in a short history |
| 37.8 to 39.15 | of software engineering. |
| 39.15 to 40.77 | So this is the Eating Act, |
| 40.77 to 44.37 | the first general purpose computer in 1945. |
| 44.37 to 47.34 | And at that time, programming was kind |
| 47.34 to 48.96 | of an afterthought, right? |
| 48.96 to 51.36 | So all that really mattered was the electrical engineering |
| 51.36 to 54.03 | stuff and trying to actually build this machine and, |
| 54.03 to 56.61 | and programming somebody else will figure that out. |
| 56.61 to 58.29 | So that's why they let the women do it. |
| 60.18 to 62.37 | In 1978 was about when I started |
| 63.54 to 65.67 | the c programming language was invented, |
| 65.67 to 69 | which is basically a high level assembly language targeted |
| 69 to 71.37 | at the PDP 11 so that they could write Unix. |
| 72.42 to 75.18 | By 2014, we kind of flipped things around |
| 75.18 to 77.82 | and said, you know, maybe we should stop focusing |
| 77.82 to 80.315 | so much on making the hardware more efficient, |
| 80.315 to 82.83 | and maybe we should focus on making the programmer more |
| 82.83 to 86.82 | efficient so we got a much nicer tool chain |
| 86.82 to 90.48 | that's maybe not as efficient, |
| 90.48 to 93.93 | but it makes better use of human effort. |
| 94.89 to 99.06 | And now in 2023, we have headlines like, |
| 99.06 to 102.45 | are these coding jobs going to exist in three years? |
| 102.45 to 105.42 | So first I wanna say don't panic. |
| 105.42 to 106.71 | I agree with Eric, |
| 106.71 to 109.47 | there's not gonna be massive unemployment. |
| 109.47 to 112.59 | If you are a software engineer, you're gonna be fine. |
| 112.59 to 115.5 | And if you're not, you're gonna be closer to one. |
| 115.5 to 117.48 | You're going to be doing more, right? |
| 117.48 to 121.83 | So I, I just had a, a talk yesterday with data science |
| 121.83 to 126.05 | who works in nature conservation, who says, you know, I, |
| 126.05 to 129.36 | I used to fiddle around a little bit with data |
| 129.36 to 131.67 | and I could make a couple plots and so on, |
| 131.67 to 133.86 | but now I can do so much more. |
| 133.86 to 138.09 | I can do things that, that are what a real programmer can do |
| 138.09 to 142.38 | because I can ask, you know, can I go in this direction? |
| 142.38 to 143.85 | Can I explore this data in this way? |
| 143.85 to 147.06 | And I can do things I never did before. So that's exciting. |
| 148.11 to 150.96 | So let's take a, a step back |
| 150.96 to 153.27 | and look more carefully at this history, right? |
| 153.27 to 155.07 | So we went from optimizing hardware |
| 155.07 to 158.25 | to optimizing the programmer's effort. |
| 158.25 to 161.315 | And in, in the future or, or, or now. |
| 161.315 to 162.93 | And going forward, I think it'll be more |
| 162.93 to 164.43 | assisting the user, right? |
| 164.43 to 167.85 | So there won't, for a lot of the tasks, you won't have |
| 167.85 to 170.07 | to bring in a professional programmer. |
| 170.07 to 172.38 | The users will be doing it for themselves, |
| 172.38 to 174.93 | or they'll be modifying existing things. |
| 176.28 to 179.2 | We went from low level code to high level |
| 179.2 to 181.66 | - To now it's gonna be more of a dialogue. |
| 181.66 to 184.21 | We're gonna be using natural language. |
| 184.21 to 186.64 | And yes, there's gonna be a programming language |
| 186.64 to 189.97 | as an intermediary, but that may not be as important. |
| 189.97 to 192.85 | Maybe all the important stuff about a programming project |
| 192.85 to 196.33 | will be in natural language, or it'll be in, in diagrams |
| 196.33 to 200.71 | or a little sketch on the back of a, of a napkin. |
| 200.71 to 203.44 | And not necessarily in a traditional programming language. |
| 204.76 to 206.895 | In the 1970s, we had simple problems. |
| 206.895 to 211.065 | So the ENIAC was, was built to compute firing trajectories |
| 211.065 to 213.94 | so that artillery could hit the bad guys no matter |
| 213.94 to 215.98 | what the weather or wind condition was. |
| 217.69 to 221.02 | And then by the seventies we could |
| 221.02 to 224.5 | do things like build Unix, which was complex in |
| 224.5 to 226.78 | that it was a whole operating system. |
| 226.78 to 228.73 | But as you remember, the Lion's book |
| 228.73 to 232.27 | that was only like a 200 page book was the whole thing. |
| 232.27 to 234.64 | Now, by the two thousands |
| 234.64 to 237.52 | or so, we had operating systems that were millions |
| 237.52 to 240.64 | of lines rather than, than, than just thousands of lines. |
| 241.57 to 244.93 | And going forward, I think we want problems that are, are, |
| 244.93 to 246.88 | are even more complex than that. |
| 246.88 to 251.05 | These so-called wicked problems for which there is no |
| 251.05 to 253.81 | exact specification, at least with an operating system. |
| 253.81 to 256.575 | You know, if it's right or wrong with wicked problems, |
| 256.575 to 258.52 | you don't even know what the right answer is. |
| 258.52 to 262.39 | Everything changes. There is no complete specification. |
| 262.39 to 267.07 | So Brian talked about this character that you could talk to |
| 267.07 to 268.9 | and he had a great conversation about the |
| 268.9 to 270.85 | Brandenburg concertos. |
| 270.85 to 273.075 | Nobody wrote a specification that says one |
| 273.075 to 274.9 | of the things this guy has to do is be able |
| 274.9 to 277 | to talk about the Brandenburg concertos, right? |
| 277 to 279.97 | You couldn't have thought of that ahead of time. |
| 279.97 to 283.185 | And yet we're now in a position where we want programs |
| 283.185 to 286.78 | that will do things that the designers never thought of. |
| 286.78 to 289.3 | And that's brand new. We didn't have that before. |
| 291.82 to 294.4 | In the beginning it was a, it was ad hoc. |
| 294.4 to 296.83 | Then we started to say, you know, |
| 296.83 to 299.685 | maybe this computing stuff is really a mathematical |
| 299.685 to 301.63 | or a logical science. |
| 301.63 to 303.73 | And we had this whole methodology for |
| 303.73 to 305.89 | how you could prove programs, correct. |
| 305.89 to 307.33 | Of course, nobody ever did that except |
| 307.33 to 309.31 | for like the 10 line programs. |
| 309.31 to 312.28 | But we kind of knew that in theory we could do that. |
| 312.28 to 314.41 | And we wrote some tests that convinced us |
| 314.41 to 316.51 | that we were getting close to a proof. |
| 318.04 to 320.5 | Nowadays in going forward, we're kind |
| 320.5 to 321.7 | of stepping away from that. |
| 321.7 to 323.71 | We're saying it's no longer a mathematical science. |
| 323.71 to 327.1 | It's more like a natural science in several ways. |
| 327.1 to 328.39 | One is the types of |
| 328.39 to 331.51 | of problems we're solving don't have mathematical |
| 331.51 to 332.59 | specifications. |
| 332.59 to 333.94 | It's, you know, I wanna be able |
| 333.94 to 336.55 | to have a good conversation about every everything, |
| 336.55 to 339.07 | what's everything, what's a good conversation? |
| 339.07 to 341.41 | We can't mathematically define that. |
| 341.41 to 344.32 | So we kind of define it empirically. |
| 344.32 to 346.6 | The other way it's in natural science is that, you know, |
| 346.6 to 349.84 | when I started you wrote everything yourself. |
| 349.84 to 353.14 | The libraries meant you had square root and sort |
| 353.14 to 354.91 | and maybe a couple other things. |
| 354.91 to 356.835 | Right? Now what do you do? Well, you |
| 356.835 to 358.82 | - Go a bunch of packages |
| 358.82 to 361.46 | and it's like being David Attenborough |
| 361.46 to 364.25 | and you know, looking at a strange animal out there |
| 364.25 to 366.56 | and saying, ah, I gotta capture this one. |
| 366.56 to 368.69 | And oh, there's a field manual that says |
| 368.69 to 370.85 | what this strange animal does, |
| 371.81 to 375.47 | but half the time the manual's wrong and you make a call |
| 375.47 to 376.58 | and it does the wrong thing. |
| 376.58 to 379.55 | And now as a natural scientist, you have to come up |
| 379.55 to 381.74 | with your theory for this is |
| 381.74 to 383.87 | what this program does when I make this call |
| 383.87 to 385.7 | to this API, right? |
| 385.7 to 388.37 | So it's not a mathematical proof that it's, it's correct, |
| 388.37 to 389.6 | it's trial and error. |
| 389.6 to 392.18 | I ran this, it didn't work. I tried using it another way. |
| 392.18 to 394.49 | Okay, now it works. I'm not gonna worry about |
| 394.49 to 395.54 | proving it correct. |
| 395.54 to 399.14 | I'm just gonna go ahead. Used |
| 399.14 to 402.02 | to be you were completely on your own in programming. |
| 402.02 to 405.92 | Then we came up with a pretty nice set of tools |
| 405.92 to 409.94 | for doing version control, doing testing and so on. |
| 409.94 to 412.4 | And, and we had that. |
| 412.4 to 415.22 | But there was kind of a separation of here's the program |
| 415.22 to 418.7 | that you run and then here's all the tool set around it. |
| 418.7 to 420.23 | And those are two separate things. |
| 420.23 to 421.315 | And I think |
| 421.315 to 424.19 | and hope that in the future we'll bring those all together |
| 424.19 to 426.92 | and we'll talk about that a little bit more in a minute. |
| 428.48 to 430.88 | Okay, so how is this used today? |
| 430.88 to 433.94 | And I wanna distinguish basically two types of use. |
| 433.94 to 437.09 | One is essentially API lookup |
| 437.09 to 439.73 | or a smart manual |
| 439.73 to 444.145 | or auto complete is like I'm programming along and, |
| 444.145 to 447.35 | and I forget how to do this one thing and I type something |
| 447.35 to 448.76 | and then it auto completes it for me |
| 448.76 to 449.81 | and I say, yeah, that's right, |
| 449.81 to 451.76 | or no, that's not quite right and I fix it. |
| 451.76 to 454.04 | And then the second is complete problem solving |
| 454.04 to 456.08 | where you just say, here's a natural language |
| 456.08 to 457.52 | description of the problem. |
| 457.52 to 458.66 | Solve the whole thing for me |
| 458.66 to 460.34 | and I'll give you one example of each. |
| 461.66 to 465.02 | So here is an example of building a video game |
| 466.31 to 468.83 | through natural language saying, well, okay, |
| 468.83 to 471.92 | when the rockets clicked display some texts saying firing |
| 471.92 to 476.69 | thrusters and temporarily speed up by four times |
| 476.69 to 478.7 | for a quarter of a second. |
| 478.7 to 481.52 | Now this is something that somebody, you know, |
| 481.52 to 486.14 | a programmer in an introductory class could build this app, |
| 486.14 to 489.56 | this little game, but they would have to look up things, |
| 489.56 to 494.56 | they'd have to know, you know, my exposition that's |
| 494.72 to 498.47 | specified by offset left in this particular |
| 498.47 to 499.58 | JavaScript framework. |
| 500.84 to 503.12 | And once you know that you have the answer, |
| 503.12 to 505.13 | but it's easier to just say it in natural language |
| 505.13 to 508.19 | and similarly, temporarily speed up, well |
| 508.19 to 511.76 | that means calling the set timer on a function and so on. |
| 512.63 to 515.24 | So it's not doing anything that the programmer couldn't do, |
| 515.24 to 518.51 | but it's just making it faster by not having to know |
| 518.51 to 523.19 | what the API is At Google, we have this joke |
| 523.19 to 526.97 | that most of what the software engineers do is take data in |
| 526.97 to 529.91 | one format, transform it into another format |
| 529.91 to 531.895 | and ship it off to another program. |
| 531.895 to 535.145 | So that's Larry and Sergei's, proto buff moving company. |
| 535.145 to 538.74 | T-shirt is popular. Okay? |
| 538.74 to 539.905 | The second type of |
| 539.905 to 544.59 | of thing is we take a problem description here from a, |
| 544.59 to 547.92 | a programming contest and it's mostly in natural language. |
| 547.92 to 549.75 | There's a little bit of formalization |
| 549.75 to 553.59 | of here's a sample input and a sample output. |
| 553.59 to 556.53 | And we say, give me the whole solution. |
| 556.53 to 561.09 | And so this was the main example from Alpha code from, |
| 561.09 to 562.26 | geez, it seems ancient now. |
| 562.26 to 563.46 | It was probably like a year ago |
| 564.6 to 567.66 | and here's the solution that it comes up with |
| 567.66 to 569.25 | and this is correct. |
| 569.25 to 570.6 | It gets the right answer |
| 571.59 to 574.8 | and kudos for that. |
| 575.7 to 577.17 | And I saw this and I said, yeah, okay, |
| 577.17 to 578.25 | it's the right answer, |
| 578.25 to 579.725 | but there's more |
| 579.725 to 582.09 | to programming than just getting the right answer. |
| 582.09 to 585.15 | So I said, you know, if if I was doing the code review |
| 585.15 to 587.13 | for this, what would I say? |
| 587.13 to 589.5 | And it turns out there's a lot of markup going on here |
| 589.5 to 592.56 | for things that are, that aren't quite right |
| 592.56 to 594.57 | that I would not let pass. |
| 594.57 to 597.39 | One of the interesting things there in the middle, this has |
| 597.39 to 600.06 | to do with popping things off of a stack |
| 600.06 to 602.435 | and it has the two stacks, A |
| 602.435 to 603.96 | and B that makes a lot of sense. |
| 603.96 to 606.27 | And then it invents this stack called C |
| 606.27 to 611.1 | and every time it pops something off of B, it saves it on C |
| 611.94 to 614.58 | and then it never uses C anywhere else. |
| 614.58 to 618.36 | And so what's happening is this system is trained on |
| 619.35 to 621.99 | on lots of programs, some of them manipulate stacks. |
| 621.99 to 624.69 | And a lot of times when you manipulate a stack, a good thing |
| 624.69 to 626.85 | to do is to save it somewhere. |
| 626.85 to 628.23 | So it had that idea |
| 628.23 to 631.17 | and then didn't realize it didn't really need that idea. |
| 631.17 to 634.44 | And a lot of programmers do that and I don't wanna hire them |
| 637.62 to 639.96 | or maybe I wanna train them to, to do better. |
| 641.31 to 643.38 | So here's a better answer. |
| 643.38 to 646.38 | And, and Ben mentioned the, the blog post, |
| 646.38 to 648.81 | this PI twos notebook where I talk about that. |
| 649.71 to 652.565 | But essentially what's happening here is |
| 652.565 to 653.61 | it's getting things wrong. |
| 653.61 to 655.92 | And I think the problem is half the programmers are |
| 655.92 to 658.05 | below average and they're all in the training data. |
| 660.03 to 661.2 | So you shouldn't expect the |
| 661.2 to 662.58 | output to be better than average. |
| 662.58 to 665.22 | And maybe we can fine tune it to do better, but, |
| 665.22 to 666.78 | but that's a fundamental problem. |
| 666.78 to 667.95 | But I think we can do better. |
| 668.91 to 673.08 | So how can we do better and how can we do this tomorrow? |
| 673.08 to 674.4 | So I mentioned |
| 674.4 to 678.54 | that we have this whole software development life cycle in |
| 678.54 to 680.31 | which many things are going on. |
| 680.31 to 682.56 | And program development is just number three. |
| 682.56 to 685.32 | There is just one of the many parts. |
| 685.32 to 688.77 | And, and I've said some ways in which |
| 688.77 to 691.89 | generative AI could be used at each point in the cycle. |
| 693.27 to 696.87 | And today, at each point in those cycles, we have a team |
| 696.87 to 699.57 | of people and they create a bunch of documentation, some |
| 699.57 to 702.24 | of it just written documentation, some |
| 702.24 to 705.09 | of it formal things like tests and so on. |
| 706.11 to 708.24 | But we really focus on this one part, |
| 708.24 to 709.62 | the software development part. |
| 709.62 to 714.62 | And that's a mix of, of code in in or JavaScript or whatever |
| 714.91 to 717.37 | and neural networks. |
| 717.37 to 720.13 | And one of the amazing things is we can do |
| 720.13 to 721.66 | this back propagation. |
| 721.66 to 724.78 | So if there's an error in the neural network, |
| 724.78 to 727.66 | we show it more examples and it gets better. |
| 727.66 to 730.06 | And that's not true anywhere else. |
| 730.06 to 732.25 | But what I would like to see is |
| 732.25 to 735.34 | what if it was true everywhere else? |
| 735.34 to 738.765 | What if we could take everything associated |
| 738.765 to 740.35 | with the software development process, |
| 740.35 to 744.04 | all this informal stuff and back propagate through that? |
| 744.94 to 747.79 | And I've seen examples of this all the time, right? |
| 747.79 to 752.17 | So we start a new project, we get the user experience people |
| 752.17 to 755.62 | to say, you know, what's the interface gonna look like? |
| 755.62 to 757.75 | And they come up with 10 different ideas |
| 757.75 to 760.96 | and they try them out and they make these prototypes |
| 760.96 to 762.67 | and they do these paper models |
| 762.67 to 764.74 | and they do these Wizard of Oz experiments |
| 764.74 to 767.2 | and they bring people into the room with the mirror |
| 767.2 to 769.72 | behind it and they examine them using the interface |
| 769.72 to 772.36 | and at the end they write this report |
| 772.36 to 774.97 | and says, here's the best interface. |
| 774.97 to 776.44 | Hand it off to the engineers |
| 776.44 to 777.94 | and the engineers implement that. |
| 778.78 to 781.84 | And then, you know, they read the report, they say, |
| 781.84 to 784.93 | this is an awesome report, you guys did a great job. |
| 784.93 to 786.31 | And then the report goes on the shelf |
| 786.31 to 788.2 | and nobody ever looks at it again. |
| 788.2 to 791.05 | 'cause it's, it's not considered part of the program. |
| 791.05 to 793.66 | The program is the code that was built |
| 794.83 to 798.04 | and they build it, it represents the, |
| 798.04 to 800.74 | the interface faithfully, everybody's happy. |
| 800.74 to 804.49 | But over time the world inevitably changes. |
| 804.49 to 807.61 | Maybe people are using devices |
| 807.61 to 810.94 | with a different size screen than they were using before. |
| 810.94 to 815.11 | And so the choices that the user experience team made |
| 815.11 to 817.42 | that were based on assumptions, |
| 817.42 to 819.37 | those assumptions are no longer valid |
| 820.21 to 822.22 | and people are kind of nervous |
| 822.22 to 824.23 | and saying, you know, things are getting a little bit worse, |
| 824.23 to 826.21 | but I'm not sure why. |
| 826.21 to 830.77 | Now if we could have all that documentation |
| 830.77 to 833.08 | as a formal part of the system |
| 833.08 to 837.94 | and differentiate through it, now we could say this is why |
| 837.94 to 839.44 | we're having this problem. |
| 839.44 to 841 | It's, we made this assumption, |
| 841 to 843.22 | this assumption is no longer true. |
| 843.22 to 846.88 | And maybe the system could automatically correct itself |
| 846.88 to 849.55 | or maybe it would just have an alert to say, |
| 849.55 to 851.74 | here's why we're having problems, here's |
| 851.74 to 853.15 | where the difficulty is, |
| 853.15 to 856 | let's do a better job in fixing that. |
| 856 to 858.1 | So that's where I think the future |
| 858.1 to 860.29 | of software is going to be. |
| 860.29 to 862.33 | I think we can build systems like that. |
| 862.33 to 865 | I think we have everything we need to do it, |
| 866.62 to 869.32 | and I think it'll be an exciting time |
| 869.32 to 871.27 | for everyone involved, right? |
| 871.27 to 875.2 | It will mean professional programmers can do a better job |
| 875.2 to 878.38 | and, and have more pride in their work. |
| 878.38 to 881.74 | And means amateur programmers like my friend, |
| 881.74 to 883.815 | the data scientists can do more |
| 883.815 to 885.37 | and build things that are bigger |
| 885.37 to 888.07 | and more ambitious than they ever could before. |
| 888.07 to 892.07 | And even non-programmers can get into the act, right? |
| 892.07 to 895.1 | So I think about the, the small enterprises, right? |
| 895.1 to 898.7 | For, you know, we got a lot of enterprise vendors here |
| 898.7 to 899.87 | and they do awesome stuff |
| 899.87 to 901.46 | and you heard some about it, you'll, |
| 901.46 to 903.35 | you'll hear some more later on. |
| 903.35 to 904.855 | But they're sort |
| 904.855 to 908.78 | of focused more on the larger size companies in which you |
| 908.78 to 909.95 | can make an investment |
| 909.95 to 914.78 | and amortize it over hundreds of users within your company |
| 914.78 to 916.73 | that are gonna use this system. |
| 916.73 to 919.4 | What if you only had two users within your company? |
| 919.4 to 922.79 | You can't really amortize bringing in a programmer |
| 922.79 to 925.37 | to create something for those two users, |
| 925.37 to 927.23 | but maybe with a system like this, |
| 927.23 to 929.21 | they could build it themselves. |
| 929.21 to 930.8 | And that's where I think the future is. |
| 930.8 to 933.14 | And it's an exciting time and I can't wait to see it. |
| 933.14 to 936.35 | And now I'm no longer standing between you and the break. |
| 936.35 to 936.98 | So thank you. |
Your Future
With tech at the center of so much change, realize that you may very well be preparing for careers that don’t yet exist. But by studying the intersection of business and technology today, you develop a base to build upon and critical thinking skills that will help you evaluate new, emerging technologies. Think you can afford to wait on tech study, and then quickly get up to speed at a later date? Whom do you expect to have an easier time adapting and leveraging a technology like social media—today’s college students who are immersed in technology or their parents who are embarrassingly dipping their toes into the waters of TikTok? Those who put off an understanding of technology risk being left in the dust.
Consider the nontechnologists who tried to enter the technology space in the early part of this century’s tech growth. News Corp head Rupert Murdoch piloted his firm to the purchase of MySpace only to see this one-time leader lose share to rivals. Former Warner executive Terry Semel presided over Yahoo!’s malaise as Google blasted past it. Barry Diller, the man widely credited with creating the Fox Network, led InterActive Corp (IAC) in the acquisition of a slew of tech firms ranging from Expedia to Ask.com, only to break the empire up as it foundered. And Time Warner head Jerry Levin presided over the acquisition of AOL, executing what many consider to be one of the most disastrous mergers in U.S. business history. Contrast these guys against the technology-centric successes of Mark Zuckerberg (Facebook), Steve Jobs (Apple), and Sergey Brin and Larry Page (Google).
While we’ll make it abundantly clear that a focus solely on technology is a recipe for disaster, a business perspective that lacks an appreciation for tech’s role is also likely to be doomed. At this point in history, technology and business are inexorably linked, and those not trained to evaluate and make decisions in this ever-shifting space risk irrelevance, marginalization, and failure.
Key Takeaways
As technology becomes cheaper and more powerful, it pervades more industries and is becoming increasingly baked into what were once nontech functional areas.
Technology is impacting every major business discipline, including finance, accounting, marketing, operations, human resources, and the law.
Tech jobs rank among the best and highest-growth positions, and tech firms rank among the best and highest-paying firms to work for.
Information systems (IS) jobs are profoundly diverse, ranging from those that require heavy programming skills to those that are focused on design, process, project management, privacy, and strategy.
Questions and Exercises
Look at Fortune’s “Best Companies to Work For” list. How many of these firms are technology firms? Which firm would you like to work for? Is it represented on this list?
Look at Bloomberg Businessweek’s “Best Places to Start Your Career” list. Is the firm you mentioned above also on this list?
What are you considering studying? What are your short-term and long-term job goals? What role will technology play in that career path? What should you be doing to ensure that you have the skills needed to compete?
Which jobs that exist today likely won’t exist at the start of the next decade? Based on your best guess on how technology will develop, can you think of jobs and skill sets that will likely emerge as critical five and ten years from now?
Explore online resources to learn technology on your own and search for programs that encourage college students. If you are from an underrepresented group in technology (i.e., a woman or minority), search for programs that provide learning and opportunity for those seeking tech careers. Share your resources with your professor via a class wiki or other mechanism to create a common resource everyone can use to #geekup. Then tweet what you create using that hashtag!
