Google I/O and the State of Big Tech AI in 2023
twitter.com
twitter.com
The thing is, thats a big IFF. What if it turns out, the net outcome is net-negative, in reputation, in quality of response, and on effect on the bottom line?
I would expect Google to implement LLM/ML/GPT behind the scenes where it shows to have value. The intrusion into the sell cycle does not have to be there, for it to add value.
Selling it, implies believing it is a value in itself.
I think its still an open question. Its a great tool, but is google in the tool up selling business?
This is a great point I’ve only lightly expanded on in the thread, even though that was the premise behind the bravery of their strategic change.
Google had the tech for years, but chose to be conservative about deploying it, exactly for the reasons you mentioned.
Ultimately I think what changed is not just the competition heating up and forcing to make Google to make a move if only to hedge its position, but also society warming up to the idea of AI with its warts (hallucination comes first), which relatively reduces the bottom line impact should something go wrong.
Of course only time will tell, but for now promises of LLM-in-the-loop looks too much like a gold rush. It’s too attractive to ignore for most economic actors.
> I think its still an open question. Its a great tool, but is google in the tool up selling business?
Google is (was?) in the ads business. This move seems to be more in the information and infrastructure business. For them to be relevant in this space, their own tools need to be state of the art in this regard. It seems most AI players are following this bottom-up approach. ChatGPT, Midjourney on Discord, Claude on Slack, but I think they’re all playing on selling infrastructure rather than tooling. It’s just an acquisition funnel.
Amazon should totally sell "we used an AI to ELI5 on our pricing and control knob mess" because I gave up on understanding the 5,430 options in their GUI.
I don't do it now, but for my older dayjob I had to run ads as a customer of google ads, and the "how ads are displayed" logic is compellingly awful and shows definite signs of "for gods sake, soak this ad to get the $ out of the client before we end the run window" which is pretty bad. If they can tune their AI smarts to intuit what is a good placement rate and stop it spiking, I'd be there.
(Google ad placement logic is good. But it's also kind of terrible)
(I don't and never have worked for a FAANG)
This will definitely happen. I've seen a few projects pop up on HN already that implement ChatGPT-based Kubernetes operators. The same will probably happen for the other use cases you mentioned, what's unclear is when, and how much human supervision will be applied, and for how long until the AI is left to do its thing unsupervised.
LLMs are black box optimizers that seem to magically fit every problem-shaped hole in the minds of business executives and the general public alike. I am both excited and terrified at the premise of the timeline we're in.