[ttul]

OpenAI’s huge advantage is several months of training by hundreds of millions of users. Google never risked putting a chat model out into the wild for fear it would lead to embarrassing misuses - typical behavior for a giant company defending a legacy business at all costs.

OpenAI has no such legacy business to defend and so was free to take the risk.

[motoxpro]

Apparently this is a contrarian take, but Google is in no way in danger of being disrupted by OpenAI.

The scale they have means that they could reduce that "advantage" in a matter of hours if they wanted to. 100m monthly active users vs Google 4.3b daily active users.

The compute they have can train more/better models than they can, and they have been working on this for longer than OpenAI has (see their keynote in 2017, Chatbot that got the guy fired because he thought it was sentient)

Microsoft helps here, but all of those things still exist. Bing is tiny compared to Google.

Microsoft will "win" the productivity tools (Teams, Office, etc.) for the same reason Google will win search.

Happy to be quoted on this in a year or so.

[jurmous]

Since Microsoft was working internally on the raw search optimised Prometheus model since mid 2022, and Google did not show any chat results showing sources. My bet will be that Google won't even show sources below Bard results even a year from now. And this difference alone and with Microsoft already starting the first rollout of Bing Chat, I feel Bing will have gained a lot of marketshare in a year. Bing can only win and Google only lose with the current marketshare numbers.

[motoxpro]

I think thats fair but again, it's over an order of magnitude difference. https://www.oberlo.com/statistics/search-engine-market-share....

[lm28469]

Material means are but one small step in this whole process.

> Happy to be quoted on this in a year or so.

Look back at the last 5-10 years and you'll have you answer. They're too busy sabotaging their search engine, cramping ads in youtube and cancelling products. Too big to fail but not too big to stall

[motoxpro]

Totally. I think a year will show the direction but 5-10 years is also fine. AI is a feature, not a product.

[galaxytachyon]

We simply don't know how early these megacorps started to work on these things. Remember the chatbot Tay? That was 2016 and I would say it was pretty advanced. Still NLP, just not LLM (I think). Google has the data advantage, but OpenAI basically has all the AI tech Microsoft has developed and I don't think Microsoft is that far behind google in term of innovations and patents.

In fact, Microsoft has better track record in releasing and commercializing new tech. Google is good at half assing new tech and throw them in a pile of garbage later.

[unoti]

> The scale they have means that they could reduce that "advantage" in a matter of hours if they wanted to. 100m monthly active users vs Google 4.3b daily active users.

I’m not so sure they have the Datacenter capacity with GPU’s to make that happen overnight, even if they have the software. They have an advantage in the demand side of that equation, but the supply side (datacenters, capacity) is a real challenge not to be overlooked.

[Mountain_Skies]

With the current state of international relations and supply chain disruptions continuing to ripple from the pandemic, if I needed to rely on being able to quickly access a large number of GPUs or other microprocessors, I'd be very nervous right now. Not sure Google is even wake enough to see any of this.

[lern_too_spel]

Training is not the issue. Each company already has a trained model that is useful for a wide range of tasks. Inference is the problem. Nobody has the infrastructure to run these models at scale. Until they do, they have to pick and choose the most valuable places to deploy them.

[freetinker]

They have everything – except the culture to deliver, imo.

[htrp]

If their [OpenAI's] server status is any indication, they've been fully tapped out for the better part of 2 months collecting prompt training data.

Not sure how 100k google employees will be able to match that degree of training data.

[actsasbuffoon]

According to a UBS study from last week, ChatGPT is the fastest growing consumer-based web app in history: https://www.reuters.com/technology/chatgpt-sets-record-faste...

This thing is taking off like wildfire. Google has very good reason to be terrified. This kind of growth is quite likely indicative of a major disruption.

[jurmous]

Microsoft stated in an interview with The Verge that they were working with this new search optimised Prometheus model since mid 2022. Months before the release of ChatGPT and the big inflow of data from real users. So this Prometheus model was created with different purposes since it can show sources and can be updated with new information on the fly. So it is quite ahead of ChatGPT. Microsoft is taking it with the direct access of OpenAI tech to the next level.

[quaintdev]

> for fear it would lead to embarrassing misuses

It's also difficult to integrate their ad business in chat model

[AraceliHarker]

Google is not such a great ethical company, is it?