I normally would, but in this case, I simply don't know where to start. The counterargument is that LLMs just don't shift the graph on productivity remotely enough to entertain this level of doomerism.
Now, it may be that people are extrapolating comparable, but non-LLM successes in AI based on our current "AI" cultural moment, but if that's the premise, I wish that were made a little clearer.
https://chatjimmy.ai being a demo of the "burn the model to an ASIC" approach being sold by Taalas[0], an approach which they use to run Llama 3.1 8B at ~17000 tokens per second.
Not to downplay their accomplishment but Llama 3.1 8B is a terrible model. It's really outdated at this point. It's cool that they were able to accelerate a model with silicon, but it also feels wasteful since llama 8B is such a useless model?
I guess their point was to demonstrate that it's possible to bake a decently-sized model to a silicon? As with anything related to HW, I guess the lead time will be considerably larger than the software counterparts, so I guess in 1-2 years timeframe we might see something like Gemma 4 baked onto a silicon.
Yeah, I think the important part is the process to convert the model to silicon, not the actual implementation itself.
Whether it succeeds now depends a lot on the rate of improvement of model architecture. They're betting on model design and capability improvements slowing down - and then wiping the floor with everyone else with their inference economics.
I think this is the future. When models start converging at "really good" (which I think is already happening) then burning them into ASIC silicon is the natural next step.
Harnesses can keep improving with a fixed model and the throughput opens up new possibilities like doing 10x more "thinking" or exploring parallel paths and picking the best.
I can't recommend being intentionally rude so that you can get practice dealing with people who are pissed off at you. Learning how to tell when it's a bad time to strike up a conversation with a stranger will be a much greater benefit for anyone looking to meet women or even for someone just working on being friendlier/more social. There'll be no shortage of opportunities to learn how to deal with rejection even without being a pest.
As I understand, GP is recommending that folks take a risk that their drumming up a conversation might be unwelcome if they are unsure. They're not advocating for folks to harass people who are in no position to chat or who have stated as much.
People can't read minds, so I think we owe grace to the folks around us if they misread a situation and respect your wishes when you let them know that it's you're unwilling or unable to chat at the moment.
While it's great to learn social cues, it's often impossible to know whether someone is in the mood to chat.
I cannot agree with that. Do not be shy about trying, but do not be a pest either. There are many times when interruptions are not good for both sides. My 2c.
Yeah, imo, it’s nowhere near ready for 1.0. I was a big advocate for this browser but recently changed because of exactly this. That, and it’s very slow after having it running constantly, I found myself routinely quitting and re-opening it every hour or two to get normal speed back, or my RAM for that matter.
You know, feel free to keep thinking this. In my experience Grok is the best. I don’t let myself into weird groupthink that happens just because trolls took advantage of Grok’s absence of lobotomy. Kind of a superpower.
There's no way I'm trusting the current driving cohort with a third dimension. If we get flying cars and they aren't completely autonomous, I am moving to the sticks.
reply