https://blenderbot.ai/

Example question: What’s a covariance matrix?

BlenderBot: Principal component analysis is a popular technique for analyzing large datasets containing a high number of dimensions/features per observation, increasing the interpretability of data while preserving the maximum amount of information, and enabling the visualization of multidimensional data.

ChatGPT: A covariance matrix is a symmetric matrix that describes the covariance between multiple variables in a given dataset. The diagonal elements of the matrix represent the variance of each variable, and the off-diagonal elements represent the covariance between each pair of variables. The covariance matrix is used in multivariate statistics to understand the relationships between different variables and to perform statistical inference and hypothesis testing.

Do these companies fear their customers that much? Maybe that's the difference between Microsoft/OpenAI and Facebook/Google. Microsoft sells the LLM to the customer who will be more impressed by the product if it's useful. With Facebook/Google the product uses the LLM. The LLM may make them an unsuitable product if they don't click on an ad after using it.

What makes you so sure? Microsoft already has a fully productized LLM that people pay for and has in turn provided real value to both users and the company (copilot). To assert that some other company has a better one but it's... not ready yet? is odd.

I can't think of the last time Google made a new product. New as in "not a reimplementation of what another company already productized", product as in "people pay for it". I don't think they have what it takes.

Edit: my bet is Google makes a free to play version that is ad supported, as they're want to do. The question is where do the ads go. The answer, most likely, is in the responses. So instead of reading the most likely continuation of a string, you get what the highest bidder wants the most likely continuation to be. Hooray.

And kill their golden goose? Google makes more money the longer you spend on their ad-ridden search result pages. I don't doubt they'll make a question answering bot, but the real question is how they fund it (see edit above for my guess).

Someone will kill their golden goose sooner or later. It's better if they kill it themselves and silently replace it with something else.

But deploying something like that at scale not only means developing the consumer experience but also the ad placement platform that generates optimal revenue from it. It's not a thing to toss over the wall just because ChatGPT won some news cycles.

So why not keep people there by having an engaging chat bot? Could even refresh and add new context sensitive ads every time you ask something new.

Chat bots also can very naturally embed advertisement:

btw, I think you would really love to check this related product which according to many reviews can help with the problem you just asked me about.

Although clicking on them might not be quite as natural as clicking as part of search results, so still massive damage to the golden goose.

I think Google has three choices: 1) reorient their business model to be about making money by providing direct value to the user, 2) pivot to be about ad-supported entertainment rather than productivity/information, or 3) cut costs, run as a cash cow, throw off as much profit as possible before obselesence.

OpenAI will have a secure home with Microsoft, but aren't positioned to run away with anything anytime soon. They're at least as likely to be the Altavista to Google Search, the Napster to Spotify, the MySpace to Facebook, etc

Until OpenAI is generating billions in profit, I wouldn't be so quick to criticize the other big companies. Google in particular has a reputation[0] to defend - they're not going to risk their cash cow with a half-cocked LLM that spouts patent falsehoods half the time. Not to mention the false economics of losing money on every single search.

[0] - the reputation might be sullied amongst the HN crowd, but they're negligible when it comes to dominance of Google as a search engine for normies.

They are trying it, here:

https://ai.facebook.com/blog/democratizing-access-to-large-s...

Demo:

https://opt.alpa.ai/

But it's clearly not as good as GPT3.5