From the very rational "AI as a Normal Technology" (formerly "AI Snake Oil") website.
AI existential risk probabilities are (still) too unreliable to inform policyI'm going to characterize this, as always, extremely well researched & written post as, there is no data, no logic, for us to compute probabilities of "AI doom" scenarios.How speculation gets laundered through pseudo-quantification
So what does that tell us about the folks preaching these scenarios? Directors of Reality TV shows?
Next up from The Glyph, via his "Decyphering the Glyph" blog. The Glyph apparently is a very accomplished open source Python contributing software developer, Glyph Lefkowitz.
What Would A Serious AI Product Look Like?A very good read. I noticed him espousing something I mentioned (I think) a few months ago:Every “AI” tool is missing critical features that you would need if you wanted to do real work with them. What would those look like, if they existed?
No First-Person Output, No Apologies
There is no reason for a software development or research tool to use first-person language to describe itself. They should not do so. In fact they should not be allowed to do so.There is also no reason that they should ever apologize. It is a waste of everyone’s time; it’s a waste for the chatbot to generate the apology, it’s a waste for the user to read the apology, and it’s a waste for the user to respond to the apology. Yet they unfailingly do this upon every correction.
Next, from Daron Acemoglu, a 2024 Nobel laureate in economics and Institute Professor of Economics at MIT, via Project Syndicate.
Distorted Intelligence at the AI FrontierThis reminds me of Cory Doctorow's assertion - I can't find a link, so from memory - "The Luddites weren't opposed to technology. They just wanted a say in who it was for, in who benefitted from it."The real problem for the frontier AI labs is not that their models are too powerful and already “misaligned” with the goals their creators set for them. Rather, it is that the models are being trained in ways that may be leading to a type of intelligence that will become less predictable.
...
This is what I mean by distorted intelligence. If my suspicion is correct, what we are dealing with is not a model racing toward superintelligence, but a brittle house of cards that becomes more and more likely to malfunction and collapse as we demand more from it.
This is similar but different. I hadn't thought about it before, but, yes, apparently LLM training has been slanted to the business models of the frontier AI companies. I would summarize these as, create as many chatbot addicts as possible. I think we saw mention of this some posts back.
Wow, these guys need to be terminated with prejudice.
From Gary Marcus. He has been such a fighter for a vision of AI beyond "autocomplete on steroids" - which he coined.
I hate to see him drinking so much of the koolaid.
BREAKING: Florida seeks injunction against OpenAII mean seriously?The big news in the media this morning is that Jensen (“What, me worry?”) Huang has announced a platform for “AI Agent Safety”.
LLM-based agents remind me of the 1st PCs & MS-DOS: who has time or resources for security? If we worry about security, we'll never be able to ship a product!
I guess 1 advantage of being 75 YO is that I can say, yes, we have seen this bullshit before.
Here's the home page/directory for my posts on Bullshit. This is post #153.
