Comments by "miraculixx" (@miraculixxs) on "TED" channel.

Define "originally"
6
There is a library to analyze a CSV file. It works very similarly but it doesn't need half a households electricity intake to compute. That's a nice demo but not impressive. Also it is not inventing things.
2
Unpopular opinion: I think people are too easily amazed, especially at TED. Connecting machines is not a new thing. Once you have something like ChatGPT it is not surprising you can connect it to some other machine. When everyone is getting excited I think the best option is to calm down and take a step back. In a gold rush sell shovels.
2
Why is he misrepresenting the capapilities of large language models? They are NOT capable of any of these tasks he claims, unless a skilled & intelligent human being is driving it. AI is a tool, not an agent. Mark is way off. The robots he showed btw do not use AI at all. They are machines that use straight forward control theory. His papers do not show what he claims. LLMs dont have a world model.
1
It is not any more difficult than any other technology development. There is a lot of hype right now, I suggest to slow down and just check back in a year or two
1
@hiddendrifts new shiny tools are not advancement, just tools. Most of them will not survive. Look at previous times in history when new foundational tech was invented. It always led to a flurry of new applications, but most of them did not survive bc they did not actually meet a demand. Yes, there will be a lot of new applications, no there is no danger of missing out, there is plenty of time to catch up once the dust settles. My approach is to wait and see what sticks. YMMV
1
Unfortunately most blockchain people don't realize that trust in a local agent is valued by people. Only very few people are happy to trust 'the system' but prefer to trust someone they know. Hence bitcoin and blockchain in general are not for the masses.
1
Anderson sure sounds like HAL.
1