The AI dooooooom thread

[Syt]

I found these presentations from Yann LeCun interesting despite the fact that he works for Meta.  It's possible he is just talking down his competitor's books, as Meta lacks an industry leading LLMs, but he seemed on the level.

https://www.youtube.com/watch?v=ETZfkkv6V7Y&t=3090s

https://www.youtube.com/watch?v=qvNCVYkHKfg&t=3122s

Thanks.

[Razgovory]

2 key assumptions are that

Existing AI models, given sufficient computing power, can and will speed up the pace of AI research and development in an accelerating feeback loop

Existing LLM models can somehow through that process bootstrap into an AGI.

I also don't understand why it suddenly decides to kill all humans.

[Josquius]

2 key assumptions are that

Existing AI models, given sufficient computing power, can and will speed up the pace of AI research and development in an accelerating feeback loop

Existing LLM models can somehow through that process bootstrap into an AGI.

I also don't understand why it suddenly decides to kill all humans.

It's ran by Elon.

[HisMajestyBOB]

2 key assumptions are that

Existing AI models, given sufficient computing power, can and will speed up the pace of AI research and development in an accelerating feeback loop

Existing LLM models can somehow through that process bootstrap into an AGI.

I also don't understand why it suddenly decides to kill all humans.

It's ran by Elon.

Checks out. If I was regularly exposed to Elon Musk I'd want to kill all humans too.

[Razgovory]

I was worried about this AI thing, but I am comforted by the knowlege that the companies that designed it are run by men with  sober-minded ideas such as "create God and ask him for immortality.". 

[The Minsky Moment]

Well two of the leading AI companies are led by an addled ketamine addict (X) and a pathological liar (Open AI).

The Anthropic guy seems kind of balanced.  On the one hand he keeps warning about the dangers of AI, the threats to economic stability and world peace, and the need for regulations. His way of addressing that in the business is to proceed at breakneck speed in AI development and do deals with any international scumbag with cash (aka the Gulf Sheikhs).

[celedhring]

Ezra Klein had a former AI developer turned whistleblower a couple weeks ago that essentially said that we should juryrig AI datacenters with explosives for when they go full skynet, because they will. He seemed a bit unhinged, but not completely.

[Grey Fox]

It's unnecessary. None of our power sources are maintenance free.

[PJL]

Honestly, the way things are going, being ruled by AI doesn't sound much worse than what is already happening.

[Jacob]

Honestly, the way things are going, being ruled by AI doesn't sound much worse than what is already happening.

"It couldn't possibly get worse."

... it can

[Valmy]

Yeah. It can always get worse.

[Sheilbh]

Joining the Meta AI powered personalised advertising in uses I think will be sustainable, Google recent announcement is making it absolutely clear that they are going very strongly on AI powered search. I wouldn't be surprised if we see it become the default and then only search on Google within a few years.

As I mentioned before publishers are in a really shitty position on this. Google are very robust on their line that if you want to be indexed for search they will scrape your site and they'll use it for indexing and training their AI - there's very little opportunity for an a la carte option. And all publishers are reporting a very strong fall in click through to underlying content from AI summaries and search results.

[Jacob]

You think it will be sustainable?

[Zoupa]

Honestly, the way things are going, being ruled by AI doesn't sound much worse than what is already happening.

"It couldn't possibly get worse."

... it can

[Sheilbh]

You think it will be sustainable?

It will be Google's search engine - that's the core of their latest message. At the minute they're summarising text and providing citations. That's significantly affecting publishers but, so far, seems to be increasingly accepted by users. The next stage on what they're announcing will be to replace the entire index of Google search with AI powered interactive experiences. I'm sure there will still be links in that experience - based on publisher experience, for a while users were initially unsure/untrusting and clicking through to the underlying source, now the number actually clicking through has collapsed.

I'd then add that if you get in on the AI personalised advertising you may end up never really needing to leave Google and your interactions powering more effective personalised advertising (with Google acting as creative agency, ad exchange, ad server and publisher).

Wouldn't be surprised to see something similar with Apple as I think AI will hugely help all of the platforms deliver the walled garden versions of the internet they've been building all along.