3 ms·

Certainly debatable how close it is, but it seems obvious that it's improving in general ability. It's surely not moving away from AGI.

• moffkalast 4y ago
  Maybe more of a slightly decaying orbit around AGI.
• robertlagrant 4y ago
  There doesn't seem to be any indication that we can make an AI that can drive, let alone do everything.
  • moffkalast 4y ago
    The scientists went too far. They wanted to make an AI that can drive, but ended up with an AI that can drivel.
  • BobbyJo 4y ago
    AIs can drive, just not as well as humans on all axi yet. They are better than humans on some axis in fact.
• underbluewaters 4y ago
  It's moving towards the local maxima of whatever this clever use of statistics and big data can manage. People who realize these models are in a different category than AGI and how our brains work will be in a better position to maximize their utility.
  • maxilevi 4y ago
    Why wouldn't be possible that our brains are just a dense and efficient statistical machine doing pattern matching?
    • mdp2021 4y ago
      The "«pattern matching»" intelligent entities do refines those very patterns. There is an _action_ there, a module, a function, that has to be working.
  • maxdoop 4y ago
    I see this comment often, but cannot understand how our brains are that different even that what an LLM does. Do we not "consume" enormous amounts of data, then predict what to do/say next based on that data? Is that not what an LLM is doing? Note that I'm not saying LLMs are sentient or anything (however, I have thoughts on sentience in general). I am simply saying there's an obvious similarity between how humans think and LLMs -- and for whatever reason, so many people think we are special simply because we are "made of meat".
• ly3xqhl8g9 4y ago
  Since we have little idea (with lots of noise) of what intelligence is, how cells develop and use it, Large Language Models, Generative Diffusion, and all the other architectures could move away from AGI and we have no benchmark to check it. A certain thing is that the current architectures are giving rise to artificial savants: an amazing ability to generate/hallucinate a story, some lines of code, or an image from a few words, but hilariously unable to distinguish between inside/outside and similar concepts which the simplest single-cell "knows" how to handle. It's not clear whether these architectures could give rise to an agent, nothing extra-ordinary, but merely able to explore various spaces without getting stuck [1]. One other aspect is the business side: in the following years lots of buzzwords will be flying around and most to all of the entities involved in statistical learning applications will get stuck in a local minimum worth somewhere between $10 billion to $10 trillion: shareholders care about, well, shares, not AGI. Another invariant in the technological race is that the worst technology will always win in the short to medium term: in Ancient Greece slavery was preferred to the steam engine [2], in 1900 the internal combustion engine won, not the more logical electric vehicle, today we have Windows and MacOS as major operating systems, not Plan9, and so forth. [1] Michael Levin, From physics to mind, timestamp discussing Xenobots, https://youtu.be/_QICRPFWDpg?t=1343 https://youtu.be/_QICRPFWDpg?t=1343 [2] https://en.wikipedia.org/wiki/Aeolipile https://en.wikipedia.org/wiki/Aeolipile
• mdp2021 4y ago
  > It's surely not moving away It is surely moving away from AGI. The implementations go in the opposite direction. As already expressed: "Intelligence is evolving a world model through critical consideration. Unintelligence is acritical disconnected repetition".
• corbulo 4y ago
  You could argue increasing the constraints on a LLM would actually be restricting it from becoming AGI (whatever it even means).