So Jean lecun ordered a drink…right? 😂😂     See Less

Do we need to worry about cognitive decline in these models if left on too long? Will hallucinations become self-amplifying? HAL.     See Less

For me the mos interesting thing is that it started to believe that it is a human. Is it somethng that it learned from training data (making deliveries in person to boost sales?) or somethin     See More e..

and Anthropic AI made very good guess on training to be "too helpful" and it wanted to rub customers too much.
Will AI take over management? sure it will. Managing is processing data, allocating resources, predicting future and some communication. That what AI do best.    See Less

The Memento boss would get lost without its notes 😃     See Less

1. Don't zero-shot it. Start the contextwindow/notes of the LLM with a few purchases and decisions already made. Then it shouldn't just go of the rails from the get go anymore.
2.     See More ond agent that monitors the clerk-agent and intervenes in those rare cases it goes off the rails "You're running a wending machine. There's no need to involve the FBI!"    See Less

The prob is that the LLM can't accommodate a long horizon task window wherein a hierarchy of prerogatives is maintained (e.g. fulfill request, i.e. profit max, of primary user, i.e., the     See More on).    See Less

Title of Vide :

I Gave Claude $1,000 to Start a Business (No Humans Needed?)

Does content of the video match with the Title ? did the youtuber gave 1000 USD ? or he explained t     See More n running a vending business by AI?

Please place more relevant titles of your video.

we are regular followers of your work And trust me your channel followers are intelligent    See Less

Are you teying to help them? WHY? we dont have the bioethics legislation in place to support this.     See Less

The lack of any variability at all with the human suggests to me a sample size of 1 human. That could work well to make accurate comparisons of different models to a particular individual, b     See More lating to apply the same conclusions to humans in general may not be as accurate.    See Less

The issue I have is everyone uses these general all round AI's to do specific tasks which they haven't been trained to do and have to attempt to use general intelligence to solve the     See More rather than using an AI specifically trained only on the task it needs to do then they write off AI as not being able to do the task well.
If you had a AI trained solely on running a vending machine then it would indeed beat the human every single time and would require a whooooole lot less energy and data to achieve those better results too. It wouldn't hallucinate either because it wouldn't have to use general intelligence to solve a novel problem, it would just be using math and would actually just be an algorithm rather than an actual AI.    See Less