Our early-adopters index examines how corporate America is deploying AI

From The Economist:

Technology stocks are having a bumper year. Despite a recent wobble, the share price of the Big Five—Alphabet, Amazon, Apple, Meta and Microsoft—has jumped by 60% since January, when measured in an equally weighted basket (see chart 1). The price of shares in one big chipmaker, Nvidia, has tripled and in another, amd, almost doubled. Their price-to-earnings ratios (which measures how much the markets think a company is worth relative to its profits) are ten times that of the median firm in the s&p 500.

The main reason for the surge is the promise of artificial intelligence (ai). Since the launch in November of Chatgpt, an ai-powered chatbot, investors have grown ever more excited about a new wave the technology that can create human-like content, from poems and chunks of code to video footage. This “generative ai” relies on large-language models which are “trained” on big chunks of the internet. Many think the technology could reshape whole industries and have as much impact on business and society as smartphones or cloud computing. Firms that can make the best use of the technology, the thinking goes, will be able to expand profit margins and gain market share.

Corporate bosses are at pains to demonstrate how they are adopting ai. On April 4th Jamie Dimon, JPMorgan Chase’s boss, said his bank had 600 machine-learning engineers and had put ai to work on more than 300 different internal applications. David Ricks, the boss of Eli Lilly, has said that the pharmaceutical giant has more than 100 projects on the go using ai. Howard Schultz, who recently stood down as boss of Starbucks, had said he planned to invest $1bn to use ai to figure out how to make the perfect vegan breakfast sandwich to accompany the coffee the firm sells.

Company case studies reveal only part of the picture. To get a broader sense of which companies and industries are adopting ai The Economist examined data on all the firms in the s&p 500. We looked at five measures: the share of issued patents that mention ai; venture-capital (vc) activity targeting ai firms; acquisitions of ai firms; job listings citing ai; and mentions of the technology on earnings calls. Because other types of ai could bring benefits for business, our analysis captures activity for all ai, not just the generative wave. The results show that even beyond tech firms the interest in ai is widespread and growing fast. Moreover, clear leaders and laggards are already emerging.

Start with the growing interest. ai expertise already seems to be spreading widely. About two-thirds of the firms in our universe have placed a job ad mentioning ai skills in the past three years says PredictLeads, a research firm. Of those that did, today 5.3% of their listed vacancies mention ai, up from a three-year average of 2.5%. In some industries the rise is more dramatic (see chart 2). In retail firms that share has jumped from 3% to 11%, while among chipmakers that proportion grew from 9% to 19%.

. . . .

The number of ai-related patents trended up between 2020 and 2022, based on data provided by Amit Seru of Stanford University. PitchBook, another research firm, concludes that in 2023 some 25% of venture deals by s&p 500 firms involved ai startups, up from 19% in 2021. GlobalData, also a research firm, finds that about half the firms scrutinised have talked about ai in their earnings calls since 2021 and that in the first quarter of this year the number of times ai was mentioned in the earnings calls of America Inc more than doubled compared with the previous quarter. Roughly half been granted a patent relating to the technology between 2020 and 2022.

The use of generative ai may eventually become even more common that other sorts of ai. That is because it is good at lots of tasks essential to running a firm. A report by McKinsey, a consultancy, argues that three-quarters of the expected value created by generative ai will come in four business functions—research and development, software engineering, marketing and customer service. To some extent, all these operations are at the core of most big businesses. Moreover, any large company with internal databases used to guide employees could find a use for an ai-powered chatbot. Morgan Stanley, a bank, is building an ai assistant that will help its wealth managers find and summarise answers from a huge internal database. slb, an oil-services company, has built a similar assistant to help service engineers.

While the adoption of ai is happening among many firms some are more enthusiastic than others. Ranking all the companies using each metric and then taking an average produces a simple scoring system. Those at the top seem to be winning over investors. Since the start of the year, the median share price of the top 100 has risen by 11% while for the lowest-scoring quintile it has not moved at all.

Link to the rest at The Economist

21 thoughts on “Our early-adopters index examines how corporate America is deploying AI”

  1. Two problems I see in their analysis:

    First is most obvious: there are different forms of “ai” with different purposes, levels of maturity, and likely returns.

    Second, is that tracking interest, investment levels, and paper outputs, says nothing about the economic significance of the output, even in the trendy generative sub-sector. For example, as of October 2022, before ChatGPT exploded into hypedom, Facebook has thrown $36B into the then trendy metaverse with nothing to show for it, not even trendiness anymore.


    There will no doubt be many attempts to quantify the prospects and current value of “AI” but since it’s not only not a single well defined thing, it’s not even a product or a market. It’s really more of a feature, a new programing technique for processing Big Data within software. “AI” in and of itself has minimal economic value, no more than any specific programming value. The value comes from the products that use it.

    The impact will be massive but it will be spread out across all the different applications that adopt it. And those will be pretty much everything that relies on databases to process any kind of input: numerical, textual, audiovisual, or interactive.

    How is the corporate world deploying “AI”? Easy: everywhere.
    The adoption of “AI” is akin to the adoption of the first high level programing languages or the move from mainframes to microcomputers.
    As the old BaSF ads said, “…it makes the things we (might) use better.”

  2. While I agree that while there will be important economic uses (and a lot of wasted attempts at utility) for this technology, the mere fact of capturing the “spend/interest” as a metric is insufficient as a guide for investment (as in the Zuckerberg/Metaverse example).

    The corporate “text/data analysis production-assist” AI promise is genuinely useful in a stand-alone function (with a promise of economic impact that justifies enthusiasm).

    I am, however, particularly fascinated by concepts such as this (from another article PV posted today):

    “[So] what if you took a character from a book, and you gave it [ChatGPT] the entire book and said, I want to just talk to this character, and you do that as an advertising campaign. And it could read the book and act as the character.”

    Now that’s a useful creative-adjacent prompt for generative AI that I would consider looking at (and then modifying the results of for eventual use). In some ways, that’s no different from using MidJourney to play with concepts for illustrating a blog post, and then being sparked by the intermediate results to change the content of the post (and the image prompt) because of something that the AI presented unexpectedly. Generative AI as a prompt for human creativity is, I believe, a real asset that’s coming into focus.

    The big disaster for generative AI, however, will be the autofeedback of AI-generated output as input into later database feeds. Unless the AI feeds can be kept “purely” human in origin, the contamination of lower quality data will destroy their utility. For a limited universe (“read this book, then talk like a character”), that’s not a problem. For a broad universe, for general purpose generative AI, that’s a disaster. Even for a corporate-bound universe (all of our marketing campaign results in the past — content and results), keeping the contamination out will be a real challenge.

    • Are you sure that’s a problem? 😉
      I see it as an education for lazy/trendy advertisers.

  3. Well, yes, 🙂 but it’s evidence that the horse is out of the barn, and there’s no putting it back in. I might not care about that particular horse, but a whole barnful just out of view headed for the fence means that the serious expected flaw in all-things-AI that belongs to the autofeedback threat has already been demonstrated. [Not that this seems to be getting a lot of attention…]

    • Welcome to the next internet arms race: AI vs AI.
      What AI does, AI can identify and counter.

      It’ll be fun to watch as long as we stay out of the crossfire.

      • True, dat.

        I hadn’t bothered before, but it took me about 30 seconds of signing up for Chatgpt to see the current limitations as a consumer for creative-adjacent use.

        1) No front end yet for self-contained universes (or add-ons of self-contained universes, such as a book series) for non-programmer consumers. (That I know of…)

        2) I fed it the prompt of “Karen Myers author of books” and got back a hodgepodge of information, better organized (like a Wikipedia article) than an equivalent list from a Google search, with some useful data (almost all of which came from sources identifiable to me) but also some very odd and incorrect data which had only a tenuous association with reality — the so-called “lies” or “hallucinations” — which were rather misleading. I could see that they came from some form of AI extrapolation based on limited context, but they were speculative and wrong.

        I found that a very illuminating exercise, since I had complete knowledge of the original material — it painted a vivid picture of the sort of confabulation that one would have to beware of in relying on it for actual facts.

        The search engines of Google et alia have their issues, but the lies they produce are at second hand (in their sources), not in their processes (bias filters/censorship excepted).

  4. I was pleasantly surprised at how easy it is to make stock art on MidJourney. But the thing I actually am interested in was a video game company mentioning that they’re interested in candidates who are interested in AI.

    I care, because I remember how much fun “AI” made gameplay in games like “Half-Life” and “Thief: the Dark Project.” Guards would react if they found the body of another guard you killed. Players could strategize how to approach fights: “Just run circles around the paladin in Neverwinter Nights 2. Her AI can’t handle it, and you’ll easily beat her.”

    So now I want to see how games in the future will take advantage of this next generation of AI. Half the fun of RPGs is thinking of how to approach a fight, especially a boss fight. What if Light of Heavens (the paladin in NWN2) could react to the player running around her? What if she did something interesting in response? What if she could be programmed to fight like Inigo Montoya or Zorro or even Luke Skywalker?

    And I really want to know what players may be able to do if any of those hypothetical games come with toolsets to make your own DLCs / mods. In the NWN2 toolset I once had to come up with a workaround to keep a monster from killing a prisoner it encountered on it’s dungeon patrol. The limitation of having the prisoner and monster being in different factions meant the monster would always slay the prisoner before the PC (player character) could speak to her. The game devs had to give another NPC an evil alignment to prevent him from being killed in a cut scene (he’s not evil in-story). But I want to see how they can use the new AI to enrich a game.

    Yes, I gather Meta sucks. *Shrug*. But I bet XBox, Sony, Obsidian, Bethesda, etc. will make something cool in the future with this new and improved tech. They better.

    Suddenly, AI just looks much more interesting than it did 5 minutes ago. Otherwise, carry on, y’all 🙂

    • I expect that part of the delay for ELDER SCROLLS 6 and FALLOUT 5 is adding NPC chatbots with *memory*. The biggest problems with the various image generation “free” apps are lack of reproducibility and worse, lack of memory and follow up. I expect those things are coming but they’re not here yet.

      Games and especially RPGs can really benefit from imoroved immersion from generative NPC code and the japanese-style “make life hell for the player” games can keep ttack of player tactic to make them even more unplayable for anybody but masochists. 😉

      Word is STARFIELD’s procedurally generated planets *will* have memory. Different players will find different planets at the same location but thecsame player will always find the same environment and creatures. Promising.

      NPCs you can converse with and who will remember your history with them will make the places and characters more immersive. And if the same code can be reused with different characters but with different settings it might even beable to run local, which most current gen “AI” apps can’t do.

      A problem for the Windows 12 era.

      Current apps not only run on the cloud but are really savants. Great at some things, stupidly dense at others.

      I’d like to be able to generate an image and then fine tune it, keeping everything the same except what I tell it to change. “Make the car blue instead of puke green.” “Move viewpoint to center the car.” Instead, the apps generateva different image from the same prompt every time. Often very different, even with no change. Too much randomness.

      On specific case: the “railings” and security rules are brain dead stupid on Bing Create. You can tell it you want an image of a towering female Space Force officer and it’ll accept it. (It won’t look anything like the USSF dress uniform but it will look military at least.) Tell it to do the same but with “very tall” and it’ll choke and threaten to ban you as if you’d said “naked”.

      One trick I’ve latched on to is to have it generate one thing only against a solid background. Then I can go to the paint program, make it transparent, and layer the elements. Works okay given the purpose: experimenting.

      It’s early times. We’re still at the chess-playing dog stage.

      • NPCs who remember interactions is definitely one of the advances I’m looking forward to; I can imagine so many possibilities for this feature. I once toyed with an idea that would rely upon that feature back when I was fiddling with the NWN2 and Dragon Age toolsets. I’ve never played the Elder Scrolls, but if they can pull this off I’m in. In the RPG arena I may consider any game that offers a look at the advanced AI capabilities a must-buy.

        What are your trusted sources on game news? I used to keep track of games better, but when all of my old sources did the “Gamers are dead” journ-o-list maneuver I quit paying attention to gamer media. I need to look for more YouTubers; I’ve inadvertently saved a lot of money just because I don’t know about upcoming releases and never think to look until Christmas / birthdays. It’s a shame, because I used to pre-order regularly. But I hear a lot of corrupt gaming journos are getting their comeuppance, and are now worried that AI will replace them.

        How tall was very tall? I think Honor Harrington is 6’2 or something. Would they really object if I tried to create a clone of her? Or the eight-foot tall Sgt. Taura? I haven’t gotten around to Bing create, but I’m side-eyeing them now.

        • 1: My youtube news sources are console based for the most part and XBOX focused because there’s less drama from the “Sony ponies”. 😉
          IGN is good.
          Destin too.
          Brad Sams is all Microsoft all the time but he has good inside sources. He broke the XBOX Series S story before MS was ready. 🙂
          For game tech analysis of consoles and PC Digital Forge is all about the tech.
          For RPGs, MrMattyPlays is all about RPGs.
          (Obsidian recently dropped a documentary about their 20 year history. The sort of thing you can do when you have MS money. A recruitment/marketing tool, I gather.)
          Then there is a whole range of community-focused channels. And the single game channels. Fallout 76 is a cottage industry all itself. On the XBOX side, PLUME NETWORK and Rand Al Thor 19 are watchable is occasional hyper.
          SPAWN WAVE is platform agnostic, technically, but it leans heavy to Nintendo and Playstation.

          2: Now about Elder Scrolls: you never played a one? Tsk, Tsk, tsk. (Any of the Bethesda Fallout Games at least?)

          I got my start in modern wRPGs with Elder Scrolls 3: MORROWIND. (It’s what got me into console gaming because a friend suggested I look into it. The original XBOX at $300 was way cheaper than the cheapest video card upgrade needed to play on PC. The same has been true of all Bethesda RPGs since. A year and a half later I finally bought another XBOX game. The ACTIVISION X-MEN games were brilliant.)

          BETHESDA requires a small book to explain in detail (and their lore and back story includes over a thousand in-game books that are in of themselves short fantasy stories. Open world, moddable, entire worlds. They are the premier E.D.E. games: explore, discover, experiment. More, the NPCs don’t have real memory but the game world does. You clear a dungeon it stays cleared. Unless a different faction moves in. My friend joined the Morag Tong assassins guild and proceeded to methodically assassinate every non-essential NPC in the MORROWIND island continent. And it stuck. I once got stuck flying in a dust storm and spent a half hour trying to fly back to safe ground. Mind you, it’s a 20+ year old game. Oblivion is 17. SKYRIM is 12 and, well, SKYRIM has been rereleased a half dozen times and it keeps selling. It has been updated for tech and currently runs 4K/60fps and better on PC. Latest version comes with all the DLC and santioned community mods. Hunt, fish, adopt waifs, build houses, and oh yes, fight dragons. Enchant weapons, create potions… Alas, no flying. But there’s mods for that. 😀


          Twelve year and people are still playing it. IGN has a bunch of video reviews for each review. 10/10 for the original.
          DRAGON AGE is great (Origins and Inquisition, especially) but SKYRIM is something else. A couple bilion bucks good. That’s why STARFIELD is so hyped up. “Skyrim in space! Yay!”)

          3: I gather you’re on PC so my best suggestion is get a month of Game Pass PC ($10) and try them. All of them. They’re all up there, from ARENA and DAGGERFALL (isometric, PC only) to SKYRIM. Ditto for the FALLOUT and WASTELAND series and the PILLARS games. Which is where AVOWED comes in, next year.

          Anyway, enough shilling.

          4: Bing Create is a WIP but free. You get 100 tries a week and if you have an MS account and use BING or other MS products you rack uo rewards points to buy stuff, which includes create tries. Images are 1024×1024 and downloadable.

          It’s actually pretty good (7/10) if you don’t insist on getting stuff exactly your way.
          I told it I wanted a man and woman looking at the moon out of a knee high picture window in a spaceship in a high orbit and it got it, first try. Really good moon, too. Tried iterating on lighting, vIew angle, etc. No problem. I was aiming for a 6ft 6 lady next to a 6 ft guy. Hardly exotic. But no, “rules violation” is the only message for very very tall. Or just very. (shrug) Very WIP. And they say so.

          I suspect that since it uses so much real world data it is hung up on average heights.

          Like I said, breaking the scene into elements for compositing helps a lot.

          Early days. And it’s not as if I have an immediate need for these tools. Just exploring and experimenting.

          • Okay, so here’s the thing about Morrowind: I actually own it on disc. But back then I had just finished playing KOTOR 1 & 2, and when I fired up Morrowind I was shocked by how drab it looked in comparison. Then, there was the fact that I could clearly see my PC’s sword striking the mooks that were attacking, yet the computer said I never hit them at all. And I said, “Nah, I’m just going to replay KOTOR 2.”

            And I never went back to see if Morrowind got better. However, the Skyrim soundtrack popped up on my YouTube recommends the other day. The music convinced me to put the game on my shopping list, so I intend to finally learn what this Elder Scrolls business is all about.

            I have been playing Fallout 3, though I wish the companions were as interesting as what you’d get in a BioWare / Obsidian game. They’re just kind of … “there.” At least the quests are fun, but the rather inert companions did make hesitant about whether Bethesda could deliver a truly immersive game.

            As for Bing and other AI art, I agree on having patience. These are the early iterations, and I expect it will get better. It’s just the game dev’s post reminded me that AI is used in video games, and it brought to mind so much fun potential for RPGs in the future. I can wait for the art AI to improve. But I’m actually more excited about gaming AI 🙂

            Back to game sites — thanks for the reminder about IGN. It has been years since I thought of them, and they were once a go-to until all game sites fell off my radar. I’ve bookmarked / subscribed to the other sites. Got a lot of catching up to do.

            • For better FALLOUT companions look to FALLOUT NEW VEGAS. Its by Obsidian.
              FALLOUT 4 also has interesting companions plus a voiced player character. STARFIELD won’t. Oddly enough, a lot of payers complained about FALOUT 4.

              As for the goal of AI NPCs, Skyrim has already been modded that way.


              It both processes your speech and replies with a good imitation of the voicecactor/actress in the basic game. A taste of what future RPGs should be like.

              MORROWIND is drab by design. It is an alien land under regular volcanic ash rain. I hated the cliff racers. Worse then mosquitos.

              I still recommend Game Pass as a try-before-you-buy (you get 10% off if you like it enough to buy) because the Bethesda games are sandboxes. The story is *your* story. You can totally ignore the main quest and the factions and just roam the countryside.

              Hint: if you do play, look up what restoration boost potions do to smithing and enchating. My main character is named Thor Odinson and he lives up to his name. His bow is named ICBM. 😉

              You have a lot of fun ahead of you, which way you go. It should tide you over until September.

        • Btw, this is a good week to test gaming news channels on youtube because of the fallout ( 😉 ) from the FTC injunction trial. Lots of meaningful discussion analysis and and each channel shows their style and biases while covering the same information.

          This is especially true because here was a “mishap”(?) with the redaction of Sony documents (they used a sharpie marker on paper and then scanned it before uploading them publicly) that allowed tech savvy reporters to image process the docs and extract juicy info. Lots of Sony proprietary data now out in the will. Some MS data but nothing damning, just a wishlist of companies they looked into maybe buying before Activision became available. Some even reinforces the MS legal position.

          Sony on the other hand is looking pretty bad and hypocritical, egging on the FTC with a narrative internal docs prove false. Not only were they lying publicly, they were shown to use profits from MS games on Playstation to pay developers *not* to develop for XBOX. It also came out that the reason MS bought BETHESDA, was because Sony was trying to pay them to make STARFIELD a Playstation only game. At which point MS basically said enough of this crap. They bought all of ZENIMAX the parent company of Bethesda and a dozen other studios and then made STARFIELD XBOX and PC only.

          Juicy stuff.

          Check a few of those channels and youtube will offer you more. As usual some will be good, some dreadful. Do avoid any slick channel with READY in the name (like, XBOX READY). Slick but pure clickbait.

          PS: Remember that in MS speak XBOX isn’t the consoles only, but the entire console, PC, and cloud ecosystem because it all runs off thensame set of development tools: DirectX. code once, run everywhere. (Also debug everywhere but that’s the nature of the beast.)

  5. I’m not sure if these things are into recursive loops. For example, suppose I asked for the root causes of the US Civil War. Then tell it to refute the first response, and then tell it to refute the second… Response X would always be a refutation of Response X-1.

    Using a single instance of the program, what kind of digital navel-gazing would result? How would results differ using two instances?

    • You can try some of it yourself. Just fire up a window or two to the chatbot of your choice.

      Both ChatGPT and Bingchat have memory but they also have a limit on their conversation length. At some point it’ll remind you it’s software and refuse to play.
      Navel gazing is going to be an expensive way to waste CPU cycles.

Comments are closed.