From deeper in the article:
"Fabian Stephany, assistant professor of AI and work at the Oxford Internet Institute, said AI might not be the bad guy it is painted as, according to CNBC. [snort!]
"'I’m really skeptical whether the layoffs that we see currently are really due to true efficiency gains. It’s rather really a projection into AI in the sense of ‘We can use AI to make good excuses,’ Stephany said."
I've been a serious cynic of this "AI" crap. All humans need are food and housing. Pretty much everything else is bullshit. In my view, the homesteading movement growing among the MAGA folks is where the action is, and "AI" is totally irrelevant.
Also, I notice that what people think only AI can do has been done with relational databases for years.
The Deep State world is falling apart, and Amazon was part of it. Sales are not down yet, but I understand that they sell mostly merchandise made in... CHINA. Do we even need all this stuff? For what we actually need, can it be made in the U.S. and sold directly from the companies to customers? The Chinese tariff thing comes and goes, but it looks to me that big tariffs would shut down Amazon. Made-in-America would do the same thing.
Also, these big organizations always have tons of dead weight when it comes to office employees that could be cleaned up. No "AI" involved.
Also, I point out that the Amazon Web Services part of the business just had a major blackout because they are badly managing servers all in one location.
Is Bezos seeing the future right now? I also recall that he has been kissing Trump's ass.
Kinda random thoughts. What do you think?
You've made a distinction without a difference. I think LLMs prove that 'a specific set of instructions' is not necessary at a fundamental level - or at the very least, it's statistically irrelevant. Given enough data to train on (and we're just getting started), context can be inferred, and often is, by LLMs.
I mean, you just just described an LLM. If someone starts shooting at my feet, I'm not going to try and start calculating ballistic trajectories, I'm going to flail. An LLM would, and could now currently, do the same.
An llm doesn't infer meaning. It is a giant nested for evaluation. As an example AI readers can't even get the context of live vs live correct. If they were "infering" they would be able understand the context and infer the correct form of the word.
And an llm in a robot body wouldn't flail, it would have had no language to evaluate from a bullet, plus it wouldnt dance from the instruction unless it was programmed to. The best an llm could do is respond with language to the instruction to dance.
Without the additional directions, an llm is a chat bot. A programmer has to write the code to do whatever its going to do when it encounters words and instructions.
You're in the weeds, and I suggest you read up even the very consumer level usages Model Context Protocols. To simplify, I can wire up a servo to a light switch, expose that via a 3 line MCP, and tell an LLM it has access to said switch. I ask it to turn on the light, it will. Extrapolate. This is here now. We use it on a much larger scale, every day.
edit: and forget programming. We train models. They can be LLMs, SMLs, CV models, etc. They can and do heuristically derive context. That's how the whole thing works.