airspresso
Joined 123 karma
- airspresso parentYou’re ignoring Jevons paradox. Everyone, both people and companies, will be making exponentially more software with these tools. Software that both needs to get created, debugged and updated to realize the intention of it. That’s what our time will be spent on as programmers.
- Apple has a solid hardware business and massive profits from their App Store tax, they are not dependent on ad business in the way Google is. Very different incentives.
- Sounds neat but what kind of range limits would that impose on each trip? Switching from one means of transportation to another, even if both are buses, increases the total travel time significantly. Not to mention all the hassle involved for passengers.
- This 100%. It should be seen as critical infrastructure because of everything it can enable when run well.
- That's how I read it XD "oh no, RL is dead too"
- Sounds like you're fighting the weights. What would it take to align the setup with what the LLM expects?
- Hard disagree. Python is so simple anyone can get up and running with coding in a few lines in the REPL.
- Hallucinations and output quality are two different problems. Hallucinations are usually expressed in perfect sounding sentences by the LLM, that's what makes it so convincing to end users.
- Yes, that's quite common.
- I agree. And I also know how much of that experience comes from having a legal dept. that are collaborative and supportive of what the tech org wants to do. Which I suspect is quite rare.
- > I keep hearing that LLMs are trained on "Internet crap" but is it true?
Karpathy repeated this in a recent interview [0], that if you'd look at random samples in the pretraining set you'd mostly see a lot of garbage text. And that it's very surprising it works at all.
The labs have focused a lot more on finetuning (posttraining) and RL lately, and from my understanding that's where all the desirable properties of an LLM are trained into it. Pretraining just teaches the LLM the semantic relations it needs as the foundation for finetuning to work.
- Nice unlock to hyperinflate their way to $100B. I'd buy an AGI spoon but preferably before hyperinflation hits. I'd expect forks to outcompete the spoons though.
- That's just a privately owned tech company then. Lots of companies never IPO.
- Definitely a choice to give it low memory bandwidth. Probably to avoid customers thinking it can replace any data center GPU for inference use-cases.
- How is that different from Google Chrome? Not debating that it's scary to have all that context collected in one place, just saying it's already been status quo on the web for years.
- In the tech industry, open source has clearly won. You're right that most end users don't particularly care. The engineers building solutions definitely care, and prefer to build on top of open source dependencies.
- That's clinging to straws. Many of us still prefer to call it Twitter since the new name makes no sense, especially when placed in a sentence.
- Man, this hits hard. I've done so much to protect my part of the org chart from the whims of others and the cost-cutting pressure of the organization at large. My team are happy. Personally, I'm burned out to the level that nothing excites me any more and it's really hard to muster the energy to even do what's needed at the job, let alone drive vision and the team forward.
- Cutting Google accounts out of your life, however, is an entirely different undertaking that would take much longer and have a big impact on how you use the web.
- Data provinence would be neat and a big benefit. But any solution that requires virtually all content publishers to change approach (here: add signing steps to their publishing workflow) is doomed to fail. There is no alternative way to do this than what OP is doing, which is to try to filter the fire hose of content into real vs not.
- Exactly. US is the outlier vs the rest of the world when it comes to car size.
- In the Nordics almost everything that gets passed as law has been thorough studies of impact and consequence first. Takes a long time but means the law has a chance of actually having the intended effect.
- Keep in mind that these compensation packages are mostly stock that doesn't unlock for years, so no, can't buy an AI compute rig today with that.
- > The only thing keeping Anthropic in business is geopolitics. If China were allowed full access to GPUs, they would probably die.
Disagree. Anthropic have a unique approach to how they post-train their models and tune it to be the way they want it. No other lab has managed to reproduce the style and personality of Claude yet, which is currently a key reason why coders prefer it. And since post-training data is secret, it'll take other providers a lot of focused effort to get close to that.
- Framework desktop with AMD Strix Halo [1] are getting there as a viable alternative. Offering up to 96 GB of unified RAM at the moment, so still a gap up to the beefiest Mac Studio alternatives though.
- My preferred approach to avoid that outcome is to divide & conquer the problem. Ask the LLM to implement each small bit in the order you'd implement it yourself given what you know about the codebase.
- Everything has failure modes, the only difference is how hard they are to trigger
- I see what you did there
- I thought it was always DNS
- Messing up with BGP makes you feel alive, that’s for sure.
But hey, if you haven’t caused an incident yet, that just means you’re still in onboarding. Those SLA downtime budgets are there to be spent.