Fate, it seems, is not without a sense of irony
Fate, it seems, is not without a sense of irony
Alt Text: an image of Agent Smith from The Matrix with the following text superimposed, "1999 was described as being the peak of human civilization in 'The Matrix' and I laughed because that obviously wouldn't age well and then the next 25 years happened and I realized that yeah maybe the machines had a point."
When I heard that line I was like "Yeah, sure. We'll never have AI in my lifespan" and you know what? I was right.
What I wasn't expecting was for a bunch of tech bros to create an advanced chatbot and announce "Behold! We have created AI, let's have it do all of our thinking for us!" while the chatbot spits out buggy code and suggests mixing glue into your pizza sauce.
I work in the gaming industry and every week I receive emails about how AI is gonna revolutionize my job and get sent to time wasting training about how to use Figma AI or other shit like that because it's the best thing ever according to HR... and it never is obviously.
At best, it's gonna make middle managing jobs easier but for devs like me, as long as the "AI" stays out of our engines and stays into the equivalent of cooperative vision boards, it does nothing for me. Not once have I tried to use it for it to turn actually useful. It's mediocre at best and I can't believe there are game devs that actually try to code with it, can't wait to see these hot garbage products come on the market.
Gawd, me too. They've started scraping my LinkedIn recommenders to try bait me in.
For context, I work at a university. The subject was something like "xxxxxx recommends you for a company like us" implying my contact had actually been behind it, but obviously they didn't.
And obviously it reads like it was written by one of the GPTs.
Had they seen our profiles, they'd actually know what it is we do and how ridiculous recommending a chat AI is. That's sooooo beneath our knowledge and expertise. Like a random suggesting Ivermectin to Dr Faucci.
I've been enjoying Copilot quite a bit while developing, particularly for languages that I'm not familiar with. I'm not worried about it replacing me, because I very clearly use my experience and knowledge to guide it and to coax answers out of it. But when you tell it exactly what you want, it's really nice to get answers back in the development language without needing to look up syntax.
"Give me some nice warning message css" was an easy, useful one.
It's effectively a better Google search.
AI is an umbrella term that covers many things we've already had for a long time, including things like machine learning. This is not a new definition of AI, it's always been this definition.
You’re not going to achieve AI on classical computers and is simply rebranded for machine learning like how 5G was advertised to bring futuristic utopia back in 2020 only to have 4K being considered a premium feature behind paid subscriptions from 𝕏 (Twitter) to YouTube.
Quantum Computers do exist but it’s far from being on the palm of your hand.
You're confusing AI and AGI: https://en.wikipedia.org/wiki/AI_effect
AGI is what people mean, when they say "AI doesn't exist": https://en.wikipedia.org/wiki/Artificial_general_intelligence
While AI is a program that can do a task associated with human intelligence: https://en.wikipedia.org/wiki/Artificial_intelligence
AI is not supposed to be an artificial human being. AI just does a task that people associated with humans (before they readjusted the definition of intelligence after it being created).
A bot that plays chess is an AI.
It used to be that AI was AI and then when AI was coopted by the stupid they had to come up qith AGI
While this can be a valuable clarification, it ignores the plain use history of the term AI, and demands that language change for our convenience.
Laypeople have always used "AI" to mean what scientists call "AGI".
Language is weird, and tech bros suck.
You won't have general purpose true AI until it can actually think and reason, llm will never do that. At most they would be a way of interaction with an AI.
I was surprised how poorly they still did as a chatbot vs ELIZA over after 50 years of potential progress and how revered they are in certain contexts.
https://www.researchgate.net/publication/375117569_Does_GPT-4_Pass_the_Turing_Test
Given the baseline is 66% the GPT-4 results are fairly impressive
The matrix is set hundreds of years in the future though
Yea, but the "singularity" type event that caused it was way before that.
Well yeah, but the machines are already ruling over us.
I genuinely do not understand these very obviously biased comments. By the very definition of AI, we have had it for decades, and suddenly people say we don't have it? I don't get it. Do you hate LLMs so much you want to change the entire definition for AI (and move it under AGI or something)? This feels unhinged, disconnected from reality, biases so strong it looks like delusions
What is delusional is calling a token generator intelligent. These programs don't know what the input is, nor do they understand what they put out. They "know" that after this sequence of tokens, what a likely successive token is based on previously supplied data.
They understand nothing. They generate nothing new. They don't think. They are not intelligent.
They are very cool, very impressive and quite useful. But intelligent? Pffffffh
This argument pre-dates the modern LLM by several decades. When the average person thinks of AI, they think of Star Wars or any of a myriad of other works of science fiction. Most people have never heard the term in any other context and so are offended by the implied comparison (in their understanding of the word) of LLM models as being equal to Data from Star Trek.
Unless you just died or are about to, you can't really confidently make that statement.
There's no technical reason to think we won't in the next ~20-50 years. We may not, and there may be a technical reason why we can't, but the previous big technical hurdles were the amount of compute needed and that computers couldn't handle fuzzy pattern matching, but modern AI has effectively found a way of solving the pattern matching problem, and current large models like ChatGPT model more "neurons" than are in the human brain, let alone the power that will be available to them in 30 years.
There's no technical reason to think we will in the next 20-50 years, either.
Was it? I thought it was always about we haven't quite figure it out what thinking really is
I don't think that's true. Parameter counts are more akin to neural connections, and the human brain has something like 100 trillion connections.
Other than that nobody has any idea how to go about it? The things called "AI" today are not precursors to AGI. The search for strong AI is still nowhere close to any breakthroughs.