Google is faltering. After years of monogamous worship of the false god virtual assistant, the company is ramping up its AI strategy as its competitors join hands and raise pitchforks. The irony is that all of this is happening because Google thought Pitchfork cornered the market.
See, in 2017, Google researchers published the article “Attention is All You Need” introducing the concept of transformers and significantly improving the capabilities of machine learning models. You don’t need to know the technical side of it (and in fact I’m not going to teach you), but it has been highly effective and powerful; Suffice it to say that it’s the T in GPT.
You may well ask why did Google give away something this amazing for free? While large private research organizations have been criticized in the past for stifling their work, over the years the trend has been towards publication. It’s a prestige game and also a concession to the researchers themselves, who want their employers not to hide their light under a bushel. There’s also likely an element of arrogance: having invented the technology, how could Google fail to make the most of it?
The capabilities we see today in ChatGPT and other large language models did not follow immediately. It takes time to understand and take advantage of a new tool, and every big tech company has been tasked with examining what the new era of AI can provide and what it needs to do so.
There’s no doubt that Google was devoting itself to AI work just like everyone else. Over the next few years, it made serious strides in designing AI computation hardware, building useful platforms for developers to test and develop machine learning models, and tons of other things from esoteric model tweaks to more recognizable things like voice synthesis. Published papers of
But there was a problem. I’ve heard this from Google employees and others in the industry, but there’s a kind of feudal aspect to the way the company works: Getting your project under the auspices of an existing flagship product, like Maps or Assistant, is a reliable way to make money. And to get staff. And so it seems that despite amassing many of the world’s best AI researchers, their talents are relegated to the fringes of corporate strategy.
Shall we see how it turns out? Here’s an (admittedly selective) short timeline:
In 2018 they announced Google Assistant Flow, Photos (things like coloring monochrome images), a smart display with a “visual-first version of the Assistant” (have you ever seen it?), Assistant in Maps, incremental improvements to AI-assistance Show it. Google News and (to their credit) MLKit.
In 2019, a rebranded and larger Smart Display, AR Search Results, AR Maps, Google Lens update, Duplex for Web (remember Duplex?), a compressed Google Assistant that works more locally, Assistant in Waze, Driving Assistant in Mode, Live Captioning and Live Relay (speech recognition) and a project to better understand people with speech impairments.
To be sure, some of these things are great! Most, however, were just an existing thing, but with a boost from AI. Many people feel a little cringe in retrospect. You really see how big companies like Google thrive to run along with the trends.
Meanwhile, in February of that year we also had the headline: “OpenAI Made a Text Generator That’s So Good, It’s Considered Too Dangerous to Release.” That was GPT-2. Not 3, not 3.5… 2.
In 2020, Google created an AI-powered Pinterest clone, then in December fired Timnit Gebru, one of the leading voices in AI ethics, over a paper pointing out the limits and dangers of the technology.
To be fair, 2020 wasn’t a great year for a lot of people — with the notable exception of OpenAI, whose co-founder Sam Altman was fired. Individually reduce the promotion of GPT-3 Because it had grown beyond fixable levels.
2021 marks the debut of Google’s own big language model, LaMDA, though the demo didn’t really sell it. Presumably they were still thinking of a reason for its existence beyond making the assistant less prone to errors.
OpenAI started the year by showing off DALL-E, the first version of the text-to-image model that would quickly become a household name. They began to show that LLMs, through systems such as CLIP, could perform more than just language tasks, and act as an all-purpose interpretation and generation engine. (To be clear, I don’t mean “artificial general intelligence” or AGI, just that this process works for more than a preset collection of verbal commands.)
In 2022, more changes to Assistant, more smart displays, more AR in Maps and $100 million acquisition of AI-generated profile pictures. OpenAI released DALL-E 2 in April and ChatGPT in December.
At some point, I suspect that in early 2022, Google executives will open their eyes and be horrified by what they see. I’m picturing the scene in Lord of the Rings where Denathor finally sees the assembled armies of Mordor. But instead of losing their minds and being prescribed by a wizard, these frantic VPs sent emails asking why some pert startup was circling around the world leader in AI. Especially they do this after practically inventing it.
Proof of this is Imagen’s exit a month after the DALL-E 2, though like practically every other interesting AI research Google promoted, it wasn’t available for anyone to test, let alone connect to an API. Then, after Meta released Make-a-Video in September, Google responded a week later with Imagen Video. Refusion made waves for generating music, and a month later, here came MusicLM (which you may not be using).
But of course it was ChatGPT that caused the Google leadership to make a rapid transformation from concern to complete flop sweats.
It will be clear to everyone involved that this type of conversational AI was markedly different from the assistant products Google had been investing in for a decade, and In fact doing pseudo-AI of everyone (effectively just exposed to a collection of natural language APIs) Drama To. This is called existential threat.
Fate or Foresight?
Now, it was bad enough that someone else, somewhat immune to fresh takeovers, had initiated the next phase of development for the search engine, and that they did so in a highly public way that drew criticism from industry leaders ranging from had captured the imagination of everyone, even industry leaders. Avoiding technology. The real turning point of the knife came unexpectedly Microsoft,
To call Bing a “rival” to Google Search is perhaps too generous — with about 3% of global searches compared to Google’s 92%, Bing is a well-heeled gatherer. Microsoft seems to have abandoned any illusions about Bing’s ability to improve its standing, and looked outside its home for help. Whether his investment in OpenAI was preternatural foresight or fortunate coincidence, at some point it became clear that he had backed a faster horse.
Perhaps in some smoke-filled room, Satya Nadella and Sam Altman conspired to exclude Google from their new world order, but in public the conversation took the form of money, and more. Whatever the backstory, Microsoft had secured its allegiance with the innovative newcomer and with it the opportunity to work its technology where it would work best.
While we’ve seen some interesting ideas being floated about how generative AI could help with productivity, coding, and even management, they haven’t been proven yet, either due to copyright concerns or in responses to AI being “creative”. Because of the tendency to be But it was clearly very good at synthesizing information to answer almost any question, from simple factual questions to complex philosophical questions, given the proper guard rails.
Jointly search to find Microsoft’s need to innovate to move forward with a core competency of the large language model, which by good opportunity or good sense it has just signed the world’s leading manufacturer as a partner prepared in The move to integrate the latest GPT model (some call it GPT-4, but I suspect OpenAI will reserve that moniker for its own first-party model) with Bing and Edge is a tremendous hail of sorts, Its the last and best game of the search engine world.
Google, clearly upset, attempted a spoiler campaign with an empty blog post on the day Microsoft was scheduled to announce its big event, OpenAI-powered Bing. Bard, apparently the name of Google’s LaMDA-based ChatGPT competitor, was just unveiled in more exclusive fashion. Capabilities promised and no hard dates or access plans.
This attempted announcement seems to have been made so hastily that its contents were barely mentioned at Google’s “Search and AI” event two days later, and indeed it survived even that kind of fact checking. tha what you want to do. They were advertising the future of the knowledge graph. The image used to illustrate Bard contained a non-trivial error, stating that the James Webb Space Telescope “took the first pictures of a planet outside our solar system.” This is untrue, and the fact that this tortured machine intelligence got it wrong, and that no one at Google noticed or cared enough to investigate, appears to have scared off investors. .
ChatGPT certainly has issues, and in fact shortly after the rollout of Microsoft’s enhanced Bing, TechCrunch was able to get an allegedly safe and appropriate AI to correct an essay by Hitler and then vaccine disinfo written last month. republished. But these are blots on an established record that includes billions of signals and conversations to the utmost satisfaction of its users.
Google steps up and tripping up its shot, so clearly speaks to a lack of readiness even at a limited, experimental level — let alone a global rollout like the one Microsoft has already begun.
In his investor call, CEO Sundar Pichai said, “I think I see this as an opportunity to rethink and reimagine and drive drive to solve more use cases for our users. It’s early days.” , But you will see that we are bold, put things up front, get feedback and iterate and make things better. Does he sound like a guy with a plan?
It’s understandable that Google wouldn’t want to kill the golden goose by prematurely merging Search with whatever half-baked general-use LL.M. They have become experts at deploying highly specialized AI, task models that do a thing or two. But when it comes to taking a big step, their comfortable position has made them stagnant.
Is this the downfall of Google? Certainly not, it will remain the default and a spectacularly profitable, somewhat ridiculous corporation for the immediate future. But investor confidence has been shaken as it turns out that Google’s failure to innovate meaningfully over the past few years may have stemmed not from wisdom and confidence, but from reticence and pride. (The FTC and Justice also can’t help but take another shot at its advertising business.)
This turn of the worm is only in its first few degrees, and we shouldn’t guess too far when the technology in question hasn’t yet proven itself to be as valuable as everyone wants to believe. If this does not happen, not only Google, but the entire technology industry will face the consequences.