@HowToAI_ this might be useful for a compact planning module in a drone or a robot, but it's not a foundation model, not a replacement for VLAs, and not evidence that the LLM scaling paradigm is over
Our newest model, π0.7, has some interesting emergent capabilities: it can control a new robot to fold shirts for which we had no shirt folding data, figure out how to use an appliance with language-based coaching, and perform a wide range of dexterous tasks all in one model!
@Kilvesfur@KlassischeMann@bryanrbeal@godofprompt@h_thoreson what about the 3060? that's what makes this different from a cell phone with downloaded files. a decently water resistant cell phone is arguably better than a pi here, in a few ways. This only makes sense to me with the local LLM and a solid source of power.
@FluxNatural@KevinTh24552036@elonmusk@JeffDean@CMantseris you say "cars are safe with 2 cameras and a gimbal" but that's not true. They are not considered safe, especially when the system operating them is poor. what I'm saying is that learning on vision-only causes relative impairment, which by definition makes it less safe.
@FluxNatural@KevinTh24552036@elonmusk@JeffDean@CMantseris ur right, lidar isn't a true touch sensor, but camera only is even worse. it fails in the same cases our vision does. An ai driver built on messy video only data has the risk of baking in hallucinations. It will learn reality more slowly, and at lower fidelity
@FluxNatural@KevinTh24552036@elonmusk@JeffDean@CMantseris a human has an understanding of the physical world bc our neural architecture was designed by eons of touch+vision experiments. lidar replaces the touch part to create safer training, bc a camera-only system can only learn by crashing or having a safety driver say "u crashed"
@thealokverse@TheVixhal unless they get put into bodies and go to war with each other, then I think we'll see convergence. But then we're getting into multi modal stuff which isn't strictly an "LLM", but definitely still "AI"
@Luhby your article "Skyrocketing ACA premiums force enrollees to make tough decisions" has an error. You state Elizabeth pays 862 which will go to 1380, but then say "having to pay three times her current tab". 1380 is less than double the amount.
@grok@peachey_keen @Jake99044594 @LucifersTweetz@StandUpForFact "come left for me" is close enough to a command. But it's important to note that she did not ignore or refuse. Not sure if there's much to criticize in Grok's responses here, it says it was largely caused by comms and technical issues, not refusal.
@grok@DannycataPault@jaymillsbearxxx@_4Nation__@PowerpuffB23 their eyes don't look like they're directed at the bow, the focus is lower. I can see how you might think it's a pun based on context, but short of one of the creators verifying it, I think it makes more sense that they're simply annoyed at blossom