DeepMind's stuff has all been impressive.

1) where can I use it? Compare this to what’s come out of OpenAI in the last two years. It’s like science fair vs. a company making actual tools.

2) how is it helping Alphabet’s bottom line? I haven’t seen it helping here either.

Alphafold is open and seems fundamentally transformative in the science space. GPT is nice but it’s a smart meme-generator at the moment. I don’t disagree with the impact on G’s bottom line, though.

Gonna have to disagree about GPT. I’ve been using it as a tutor to learn ML on the side, and it’s literally the best tutor I’ve ever had.

Sure, I agree they are useful. My objection is it’s more in the tool category than science, while Alphafold is both. There isn’t convincing evidence that GPTs are pushing what we know; rather, they make it easier to process/search what we already know. You could hire an ML expert to be your tutor without GPTs and you’d get equal or better tutoring, though at a higher price. You can’t hire people to predict protein folding better than Alphafold. It’s very convenient that GPTs exist and they can provide tons of value, but they’re essentially the next version of mechanical turk or a domain expert you’d hire for contract work except more scalable. The net impact of GPTs may also be higher due to how often we use text, but I’d rather see a society curing disease, etc. than one generating fake books, etc.

I see what you're saying. However, I'd argue that the zero-shot learning capabilities of the latest GPTs, if continually improved upon, could potentially offer a path towards a generalist "scientist" agent, one that could perform its own research and take over R&D for humanity (aka, the singularity). But yes, I absolutely agree that the current gen capabilities of GPTs are nowhere close to this hypothetical situation.

Of course, the models from Alphabet don't really fit this bill either. I do wonder how many "protein folding" style problems there are out there, for these narrow superintelligent AIs to solve.

> where can I use it? Compare this to what’s come out of OpenAI in the last two years

If you are using OpenAI stuff you are using it.

> how is it helping Alphabet’s bottom line? I haven’t seen it helping here either.

I guess I wasn't talking about "bottom line" impressive. Oracle might make a lot by squeezing current licensees in an unimpressive way, for example.

True but they are quite independent of Google from my understanding.

No stapler for you!

That was even 15 years ago (September 2008, time flies...).

The very nature of how people search is changing. Personally, I'm LOVING phind.com, it hits all my buttons of what a modern AI inspired SE should be. IMHO someday soon, Google will just be the Youtube company. That will be their primary thing, and maybe that's good so they can make that better so it doesn't fail too.

Meh - using phind a few times so far :

  - regurgitates stuff it hits on a shallow search as authoritative response (communicating uncertainty would be a great improvement for GPT models but I'm guessing that's not going to happen because of shallow RHLF preferences)
  - search index is worse than google (eg. I've tried a search where google lands on a good solution, phind hits official docs and offers suboptimal solution)
  - produces results slower than I can read source 
  - I still need to go to the source for full reference or do follow-up (but again it's slow)

Not seeing the value tbh. If it was gpt 3.5 fast with 4 quality now we might be on to something.

I thought Microsoft was well dead, but hey…here we are.