jaycal6
Inspired
If possible, sure!
Good luck with that....
Especially since that would depend on congress the most functional body on the planet!
Exactly! Couple of Big Pharma’s Motos... Keep em’ alive just long enough so we can thrive. Never a cure or we’ll end up poor.
...and corporations, with their wonderfully high moral and ethical standards.Especially since that would depend on congress the most functional body on the planet!
Big Corps definitely, they will try to get away with whatever they can before being caught with their hand in the cookie jar (already are). But also open source would be required to follow the same regulations (if any existed). Also the corps will lobby like crazy to corrupt as many politicians as possible to get their way.
Yep, it's already out in the wild, and every AI/ML hobbyist and entrepreneur is jumping on the bandwagon to play with it or get products out there. It is getting exponentially cheaper to make AIs/LLMs.
And the big boys too.
From the article said:
It still depends on what kind of code you are writing. An efficient cluster of asynchronous Vert.x microservices? Not yet. I tried copilot and a couple of others on that one earlier this week. But I'm sure it's only a short matter of time before coding agents can nail them all.
It could be that chatbots are more producing Artificial Communication or Synthetic Text than embodying Artificial Intelligence. Applying this idea to art, we might call what they generate Artificial Art or Synthetic Art. However, "artificial intelligence" (AI, AHI, AGI, ASI) has been in the public vernacular for 50+ years, largely coming from science fiction, so to differentiate these will not happen(!), especially as companies market their gadgets as (super amazing) AI.
Yes. LLMs are simply pattern matchers. That's a part of intelligence (or, one kind of intelligence), but shy of 'general intelligence' (at least as originally conceived).
Language models in the chatbot domain help with meaning and semantics (yes based on pattern matching). But when you combine that understanding of conversational state with a domain specific neural network capable of more advanced inference and reasoning, then the capabilities go way beyond understanding conversation and text. That is where the generative AI capabilities come in. For the domains that generative AI has been trained on (imagery, music, code, fiction, poetry, etc.) - the results are scary.
Yes. It is and isn't just doing simple linear string pattern-matching and extrapolation (prediction), because it has multiple "attention" nodes which somehow allows it to do multiple levels of "pattern matching" context and concepts. It appears that it can do a human-level of conceptualization and abstraction (yes with some holes/errors), but also computer programming which requires fidelity of multiple abstractions and relationships, not just a convincing hallucination of them.
Even the pretty simplistic Auto-GPT is a scary 2nd-order implementation of GPT being able to plan and delegate tasks/planning to other GPTs and APIs to execute tasks (so far only in the computer domain). It has memory and it can be put in an infinite loop (if one is willing to pay for it) which could keep at a "problem" indefinitely, e.g. find a vulnerability in X system or robo-tweet millions of people with inflammatory misinformation asking for $$ for the cause. This isn't even looking at sending out innumerable autonomous agents to do tasks A, B, C... either working independently or coordinating or learning in the process and coordinating etc.
Yes: the underlying architecture is a neural net, which is, in essence, a pattern matcher. In the early 00s, some of these systems were used for machine vision, and there was a time in which it was unclear whether they could even properly identify objects. Times have changed.
