Google’s two-year frenzy to catch up with OpenAI

<a href="https:&#x2F;&#x2F;archive.ph&#x2F;NQqCj" rel="nofollow">https:&#x2F;&#x2F;archive.ph&#x2F;NQqCj</a>

I was skeptical at first, but I think Google’s catch up game with OpenAI has been going pretty well so far. Gemini 2.0 Pro and Flash models are really nice. Deep research feature is done really well. Context window is still the best in the industry. Integration with search, gmail, google office suite, google meet, android, etc.<p>They finally have good enough models that lets them leverage their existing portfolio of products, their cloud infrastructure and how embedded they are in the modern work life.<p>Plus, unlike Apple, they are not as restricted in their access to the training data, because of their much less principled privacy stance.

I think the biggest thing Google has working against them is the move to put lightweight models in everything for everyone. Whatever model they use for search is probably somewhere around an ~8B model, and flash 2.0 is decent but still a light weight model.<p>People now associate Google AI&#x2F;Gemini with shitty search results and bad answers.<p>Meanwhile, their SOTA models have been strong, and Gemini 2.5 looks like it might actually have taken the AI throne yesterday.[1]<p>[1]<a href="https:&#x2F;&#x2F;simonwillison.net&#x2F;2025&#x2F;Mar&#x2F;25&#x2F;gemini&#x2F;" rel="nofollow">https:&#x2F;&#x2F;simonwillison.net&#x2F;2025&#x2F;Mar&#x2F;25&#x2F;gemini&#x2F;</a>

OpenAI is being graded on a curve. They are not a public company and are not making money. Google is. That said, they definitely floundered by not productionizing transformer decoders, just like with Google meet&#x2F;Zoom. (Encoders like BERT were&#x2F;are widely used.)

When I worked at Google there was an internal link to test out &quot;Meena&quot;, their predecessor to later LLM chatbot work. This was 2019 time frame? Maybe 2020?<p>It was uncanny and creepy, pretended to be a conscious being, frequently lied, and led directly to that guy who claimed it had personhood... I can completely understand why Google chose to keep a lid on this kind of thing, hoping to be able to clean it up and produce something that could be reliably used for a product instead of a novelty. That&#x27;s back when Google was still sort-of pretending to have ethics (though they didn&#x27;t)<p>OpenAI beat them to the public presentation of this stuff because they didn&#x27;t care.

Google leadership has had a measured approach and product releases seem more baked than ever, it’s refreshing to see. Their 0 to 1s feel compelling and more Apple(2000s) like.

Google:<p>- Google AI Studio - Gemini app - Gemini app for Gemini Advanced users - Vertex AI - NotebookLM - many more I forgot…<p>vs<p>ChatGPT.com<p>This is Google’s main problem. Where several groups are trying to build the same product and compete on user attention and distract focus.<p>Search Google. Search. Show search results on the right columns with ads on the top as it is today, and the Gemini thingy on the left. It’s that simple.

As an investor (smallish): I think Alphabet&#x2F;Google can do much better with a CEO who isn&#x27;t Sundar.<p>Also: Generally: Do consider investing in companies that run services where you feel compelled to subscribe (like Youtube Premium and previously Netflix).

The problem for Google is, people who are not familiar with tech has started seeing AI as a different product than google(search)!<p>Instead of searching (google) , let me ask AI( chatgpts) and google is on the losing side of this perception war. This cannot be solved that quickly .<p>Especially with what google has done in the AI space(to a layman) it was Bard(anyone remembers) and then it was something and then its Gemini now.<p>what is the differentiator now? is google offering more free stuff than its peers? a layman doesnt care about whether it succeeds in solving a math problem or not! As long as people think these two are separate things(ai v search) , google is gonna have a problem

Has anyone met Googlers that are confident in the company&#x27;s AI strategy? Anecdotally, everyone I&#x27;ve talked to seems to have serious concerns but that might just be a small sample size.

i think it was Eric Schmidt who said that &quot;everyone would need an Assistant&quot; - about 10 years ago. (maybe in [0] ~ 2016). Something to impersonate, talking to it.<p>Why they didn&#x27;t do it, even a resemblance of - when they could.. no idea. Instead, meddling with mailboxes and the like.<p>[0] <a href="https:&#x2F;&#x2F;spectrum.ieee.org&#x2F;google-eric-schmidt-ai" rel="nofollow">https:&#x2F;&#x2F;spectrum.ieee.org&#x2F;google-eric-schmidt-ai</a>

It seems if Google had been competitive, all they would’ve managed to do is burn ~10 billion dollars in the past 2 years, and run up against the same performance wall everyone else has<p>If I were them, I’d wait until OpenAI is forced to at least slow their burn and then jump in, and undercut everyone.<p>Seems like the real money is made on the API side anyways, so having a large consumer presence isn’t super valuable yet.

I&#x27;m using both gemini pro and chatgpt plus&#x2F;pro all the time, while gemini is FAST, its answers are systematically worse. So much so that I don&#x27;t use it for anything serious anymore.<p>It&#x27;s a shame because the 1-2M context window is really amazing.

Jeff Dean has &quot;been building neural networks for decades&quot;? That doesn&#x27;t seem correct: I thought he was primarily a distributed systems researcher before becoming involved with ML?

I think ChatGPT still has a small lead, but Gemini is pretty darn good. One of the things I really like about the version I am using is the integration with Gmail, Docs, Sheets, Slides, etc.

OpenAI is already solving the problem for all its competitors by hobbling its models so much.<p>With voice mode OpenAI hears its own voice as interruptions, then misrecognizes what it itself is saying to be some topic that is off limits. It then gets into a loop of “I can’t talk about that… would you like to talk about something else?” And this is with completely innocuous topics. Absolutely infuriating. No idea how they are possibly dog fooding this and not seeing the problem.<p>The insane level of overcranking the PC knobs led me to end my subscription and add more Google to my portfolio. I’m sure I’m not alone. That’s not to say Google won’t have the same issues, but there’s always Grok.

Until Google shows they got rid of all the people responsible for stupid stuff like Black Napoleon and all the far-left politically charged garbage, I&#x27;ll pass. It&#x27;s one version mess up away from reverting to that while the same people are behind it.

We need more ai overviews because look at all the likes! ...... (word to the wise it&#x27;s a simple link traffic siphon on advertisers)<p>Kent W. said to hurry up, hmm I could only wonder why, what could he possibly see on the legal horizon.

It’s amazing how much wealth is being acquired by the people working on this. And it’s delivering what? A way to automate navigating websites or telling someone looking at the Buckingham Palace that it’s the buckingham palace ? Silicon Valley hubris continues to astound

Google screwed up so massively in this whole AI race. It blows my mind that they had the technology available internally years before OpenAI released ChatGPT but didn&#x27;t do anything with it and in fact lost top talent due to their hesitation to release it (as the article states about Noam Shazeer). And now, for basically every non-techie person, ChatGPT is synonymous with AI in the same way that Google was synonymous with search. What a disaster. Quite frankly shocked that their executive team is still in place.

2.0 Flash seems to work well&#x2F;fine for some use cases but for other produces hilariously bad results compared to ChatGPT and Claude. As a result I tend to avoid prototyping with it and instead consider it something I might implement at scale for speed and cost reasons if it turns out to work for a given use case.

Deepseek showed us that the easiest way to catch up with OpenAI is.. &lt;checks notes&gt; hiring programmers who know what they are doing.

I can&#x27;t believe that anyone in leadership at Google still has their jobs. Aside from AMD it&#x27;s the company that is wasting the most amount of talent.<p>Gemini is still garbage compared to o3-mini-high when it comes to code. GCP is still a horrible platform compared to AWS. Why would anyone pick Google models to do anything? They refusal rate is obscene.<p>Google lived too long with no strategy, just search and random noise. They don&#x27;t seem to be able to pivot. Until they do something company-wide that includes how promotions work and turning over management I can&#x27;t see them as being anything other than a 3rd rate player in AI.<p>Search will die to AI in time. And they&#x27;ve got nothing to entice people to use Google AI over anything else.