Also Alan Turing if he were alive today: “We now have access to billions of times more computing power than I did, and the world uses it for what!?”
Edit: A major flaw with Turing’s premise is that the test is not time-bound. At first blush, a conversation with an LLM is kind of miraculous compared to what we had before. But ample exposure to any LLM reveals its lack of intelligence; it just takes time.
This here is a USB charger. It uses a CPU to figure out if you connected a Samsung Phone for 50W charging, or if it should do the default 2.5W charging profile.
It’s processing power? Uhhh, about 4Million clocks/sec with 640,000 bytes of memory. Oh yeah I guess about 5-million bits.
Oh this other thing? That’s a mouse. It tells the computer if my hand moved forward or backwards or left or right. It runs a fourier transform over an infrared image and memorizes the desk. It performs a full image process / fourier transform 100 times a second to accurately track our hands and clicks. The USB connection is also a network of networks consisting of a ReedSolomon error correction code for reliable transmission at a bit over 10-million baud transfer rate.
Our real computers are doing… Porn. AI generated porn.
Fun fact: USB Chargers have more processing power and RAM than fucking the Lunar Lander / Apollo Space Program. Figuring out if Samsung phone or not-samsung Phone has so much processing power allocated to the task it’s kind of hilarious
Idk. The people I know who use LLMs the most are also the people who use human pronouns on them and think they’re conscious. It’s like the people who use them a medium amount who aren’t fooled, and I think it’s because the people who get really really into it are just all in mo matter what and just ignore any issues and the people who had a few conversations and were amazed don’t get a lot of utility out of it.
Yeah, an LLM that’s good enough at imitating human speech and writing can pass the test easily even though it’s just a program stringing words together. Nothing remotely approaching actual thought.
Also Alan Turing if he were alive today: “We now have access to billions of times more computing power than I did, and the world uses it for what!?”
Edit: A major flaw with Turing’s premise is that the test is not time-bound. At first blush, a conversation with an LLM is kind of miraculous compared to what we had before. But ample exposure to any LLM reveals its lack of intelligence; it just takes time.
This here is a USB charger. It uses a CPU to figure out if you connected a Samsung Phone for 50W charging, or if it should do the default 2.5W charging profile.
It’s processing power? Uhhh, about 4Million clocks/sec with 640,000 bytes of memory. Oh yeah I guess about 5-million bits.
Oh this other thing? That’s a mouse. It tells the computer if my hand moved forward or backwards or left or right. It runs a fourier transform over an infrared image and memorizes the desk. It performs a full image process / fourier transform 100 times a second to accurately track our hands and clicks. The USB connection is also a network of networks consisting of a ReedSolomon error correction code for reliable transmission at a bit over 10-million baud transfer rate.
Our real computers are doing… Porn. AI generated porn.
Fun fact: USB Chargers have more processing power and RAM than fucking the Lunar Lander / Apollo Space Program. Figuring out if Samsung phone or not-samsung Phone has so much processing power allocated to the task it’s kind of hilarious
And yet, we still can’t tell it’s from its.
Tracking each other. Is Tim Berners Lee even happy about the situation?
Idk. The people I know who use LLMs the most are also the people who use human pronouns on them and think they’re conscious. It’s like the people who use them a medium amount who aren’t fooled, and I think it’s because the people who get really really into it are just all in mo matter what and just ignore any issues and the people who had a few conversations and were amazed don’t get a lot of utility out of it.
Yeah, an LLM that’s good enough at imitating human speech and writing can pass the test easily even though it’s just a program stringing words together. Nothing remotely approaching actual thought.