Google falters. After years of resolute worship of the false god's virtual assistant, the company is ramping up its AI strategy as its competitors join hands and raise their pitchforks. The irony is that it's all happening because Google thought it had the fork market cornered.
See, in 2017, Google researchers published the article “All You Need Is Attention”, introducing the concept of the transformer and vastly improving the capabilities of machine learning models. You don't have to know the technical side, but it has been enormously influential and empowering; suffice it to say that it is the T in GPT.
You may well ask, why did Google give away this wonderful idea freely? While large private research teams have been criticized in the past for withholding their work, the trend in recent years has been toward publication. This is a prestigious work and also a concession to the researchers themselves, who would prefer their employer not hide their light. There is also likely an element of hubris: having invented the technology, how could Google fail to exploit it in the best way?
The capabilities we see in ChatGPT and other large language models today didn't follow immediately. It takes time to understand and take advantage of a new tool, and every major tech company went to work examining what the new age of AI could deliver and what was needed to do it.
help wizard
There's no question that Google was getting down to AI work like everyone else. Over the next several years, he made great strides in AI computing hardware design, created useful platforms for developers to test and build machine learning models on, and published tons of articles on everything from esoteric model tuning to more recognizable things like voice synthesis.
But there was a problem. Some employees at Google and others in the industry have a sort of feudal aspect to the way the company runs: bringing your project under the auspices of an existing major product, like Maps or Assistant, is a reliable way for get money and staff. And so it seems that, despite having amassed many of the world's best AI researchers, his talents were funneled into the routines of corporate strategy.
How did that turn out? It can be seen in a small timeline (admittedly selective):
En 2018, showed incremental improvements to the 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, AI assistant Google News and (to their credit) MLKit.
En 2019, a rebranded and larger smart display, AR search results, AR Maps, Google Lens updates, Duplex for the web (remember Duplex?), a compressed Google Assistant that does more locally, Assistant in Waze, Assistant in driving mode, live subtitles and live streaming (voice recognition) and a project to better understand people with speech disabilities.
No doubt some of these things are great. Most, however, were just something existing, but with a boost from the AI. Many feel a little shuddering in retrospect. You really see how big companies like Google act in thrall to trends and push them.
Meanwhile, in February of that year, the headline also appeared: "OpenAI created a text generator so good, it is considered too dangerous for release." That was GPT-2. Not 3, not 3.5… version 2.
En 2020, Google made an AI-powered clone of Pinterest, then in December fired Timnit Gebru, one of the leading voices on AI ethics, for an article pointing out the limits and dangers of the technology.
To be fair, 2020 It wasn't a great year for many people, with the notable exception of OpenAI, whose co-founder Sam Altman had to personally squash the GPT-3 hype because it had grown beyond sustainable levels.
2021 saw the debut of Google's big language model, LaMDA, although the demos didn't really sell it. Presumably they were still looking for a reason for it to exist beyond making the Assistant throw fewer errors.
OpenAI kicked off the year by showcasing DALL-E, the first version of the text-to-image model that would soon become a household name. They had begun to demonstrate that LLMs, through systems such as CLIP, it can perform more than language tasks, and acted more as an all-purpose rendering and rendering engine. To be clear, it's not "artificial general intelligence" or AGI, just that the process worked for nothing more than a preset collection of verbal commands.
En 2022, more tweaks in Assistant, more smart displays, more AR in Maps, and a $100 million acquisition of AI-generated profile pictures. OpenAI released DALL-E 2 in April and ChatGPT in December.
At some point, I suspect in early 2022, Google executives opened their eyes and what they saw scared the hell out of them. Can you imagine the scene in The Lord of the Rings, where Denethor finally watches the assembled armies of Mordor. But rather than lose their minds and be fooled by a magician, these frenzied VPs sent out emails asking why a startup was circling the world leader in AI. Especially after they pretty much invented the means to do it.
The evidence for this is the output of Image a month after DALL-E 2, though like pretty much every other interesting AI research Google publicized, it wasn't available for anyone to test, let alone plug into an API. Then after Meta released Make A Video in September, Google responded with Image Video a week later. rifusion launched to generate music, and a month later, here it comes MusicLM (which you can't use).
But surely it was ChatGPT that sent Google's leadership quickly from anxiety to outright sweat.
It would have been clear to everyone involved that this kind of conversational AI was categorically different from the Assistant products Google had been investing in for a decade, and it was in fact what everyone else's pseudo-AIs (effectively just natural language interfaces to a collection of APIs). That is what is called an existential threat.
luck or foresight
Now, it was bad enough that someone else, some takeover-immune upstart, had sparked the next phase of search engine evolution, and done so in a very public way that captured the imagination of everyone from the leaders from the industry to the non-believer in technology. The real twist of the knife came unexpectedly from Microsoft.
Calling Bing a "rival" to Google Search is perhaps a little too generous: with roughly 3% of global search compared to Google's 92%, Bing is more of a wealthy horsefly. Microsoft appears to have abandoned any illusions about Bing's ability to improve its position and looked outside its own home for help. Whether their investment in OpenAI was supernatural foresight or lucky fluke, it became clear at some point that they had bet on a fast 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 lots of it. Whatever the backstory, Microsoft had secured its loyalty to the innovative newcomer and with him the chance to put its technology to work where it would do the most good.
While we've seen some interesting ideas about how generative AI can help in productivity, coding, and even management, they haven't been tested yet, either due to copyright issues or AI's tendency to be too much." creative” in their responses. But with the right guardrails, he was clearly very good at synthesizing information to answer almost any question, from simple factual queries to complex philosophical questions.
Now they combined Microsoft's need to innovate to get ahead with a core competency of great language models, which by good chance or common sense had just lined up the world's top developer as a partner. The move to integrate the latest GPT model (some call it GPT-4, but I suspect OpenAI will reserve that moniker for its own proprietary model) with Bing and Edge is something of a forced Hail Mary, their latest and greatest move in the world of seekers.
Google, clearly uneasy, attempted a spoiler campaign with an empty blog post the day before Microsoft scheduled its big event to announce the OpenAI-powered Bing. Bard, apparently the name of Google's LaMDA-based ChatGPT competitor, was unveiled in typically sober fashion. Capacity promises and no hard dates or access plans.
This announcement attempt appears to have been done in such haste that its content was barely mentioned at Google's "Search and AI" event two days later, and in fact also escaped the kind of fact-checking you'd want to do if they announced the future. of the knowledge graph. The image used to illustrate Bard it contained a non-trivial error, saying that the James Webb Space Telescope "took the first images of a planet outside our solar system." This is not true, and the fact that this vaunted AI got it wrong, and that no one at Google noticed or cared enough to check it out, seems to have spooked investors.
ChatGPT certainly has problems, and in fact, immediately after the release of Microsoft's enhanced Bing, it was able to get the supposedly safe and appropriate AI to cobble together a Hitler essay and then regurgitate the vaccine information that an earlier version of itself wrote the last month. But these are imperfections in an established record that includes billions of directions and conversations served, to the overwhelming satisfaction of its users.
Google rushing its shot and stumbling so visibly shows a lack of preparation 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, reimagine and push Search to solve more use cases for our users as well. It's early days, but you'll see us be bold, get things out, get feedback, and iterate and improve things." Does that sound like a man with a plan?
Google understandably doesn't want to kill the cash cow by prematurely merging Search with whatever half-baked general-purpose LLMs they have in-house. They have become adept at implementing highly specialized AI task models that do one or two things. But when it comes to making a big move, their comfortable position has loaded them with momentum.
Is it the fall of Google? Of course not, it will remain the default and fabulously profitable corporation, which is ridiculous for the foreseeable future. But investor confidence has taken a hit, as it turns out that the fracaso Google's ability to innovate significantly in recent years might not have been out of wisdom and confidence, but reluctance and pride (the FTC and Justice, trying another shot at its ad business, can't help either).
However, this twist of the screw is only in its early stages, and one should not speculate too much when the technology in question has not yet proven itself as valuable as everyone wants to believe it to be. If not, the entire tech industry will face the consequences, not just Google.
