Will Kramer
@willkramer
Followers
304
Following
19K
Statuses
620
Co-founder of @laila_team - obsessed w/tech 🤓, learning 📚, and giving back 🎗
New York, USA
Joined March 2009
@ValerieAnne1970 If this is true, why are there no lawsuits? The people allegedly involved are exceptionally wealthy, so there’s a strong incentive to sue for damages. Like most extraordinary claims, something isn’t quite adding up. 🤔
5
1
9
@RandomSprint You’re worried AI will replace all lawyers? The Supreme Court won’t even allow video cameras but you think AI litigators are coming? AI will impact the lowest performers, as it should. Too many people enter law, medicine, and programming for the money and lack the aptitude.
0
0
1
@TheSouthGAJohn @unusual_whales Interesting idea, but there’s never been a conditional pardon before and the presidential pardon power has already been pushed to the boundaries enough with the “get out of jail free card for the last 10 years” thing.
1
0
4
@Adam_Karpiak When you see an odd combination of requirements like this in a JD it probably means this was written with a particular person in mind and is only posted externally because of compliance requirements.
1
0
3
If you’re using an affordable GPU, you’ll need to run a quantized model, which reduces numerical precision (e.g., from 32-bit to 8-bit) to shrink the model so it fits into GPU RAM. Running the full 670B model on GPUs would easily require $100k+ in hardware. GPUs process models much faster than CPUs, but at a significant cost. Quantization does reduce accuracy, though in many cases, the difference may not matter depending on your task. Everything is a trade off between cost/power and precision.
1
0
1
@philipsen_mark @signulll It’s a good thing they don’t actually have the money yet… @elonmusk must be tired from all this winning
0
0
1
@TechCrunch Where would OpenAI get $19 billion, given they don’t generate positive free cash flow and operate at a loss? They must be reallocating current investor funds into this initiative. This feels like another questionable move like nonprofit to for-profit 🚩
0
0
0
@galeforceVC A logical prediction. The risk/reward ratio of VC investments only makes sense for institutional investors when safer alternatives offer near-zero returns. The era of 0% interest rates significantly skewed perspectives on what “normal” investments should look like.
0
0
1
DeepSeek R1 is as good as OpenAI o1 at 3% of the cost. OpenAI’s dominance doesn’t seem likely to last, as their moat is shallow. Google dominated and won search. Why doesn’t it seem like one company will dominate in the AI era? What’s different this time?
0
0
1
@jojopetwit @protosphinx Side projects are great, but if Newton had devoted himself entirely to calculus, he might have published first and avoided sharing credit for its creation with Gottfried Leibniz 🤷🏼‍♂️
0
0
0
@rexsalisbury It doesn’t feel like a P2P app. Zelle operates on a centralized banking network, was created by banks, and requires a participating bank account to use. It’s a way for consumers to send a free, pseudo domestic wire transfer without needing the recipient’s ABA or account numbers.
0
0
1
I agree that what constitutes a “biased dataset” is inherently subjective. However, it is the responsibility of individuals to critically evaluate test scores themselves. Maintaining choice is essential to preserving a free society. Additionally, there are historical tests that are objectively binary, e.g., did the Tiananmen Square protests happen in 1989? Was the gentleman who stood in front of the tank in Tiananmen Square ever seen again in public?
2
0
0
@al_k_000 @MatthewBerman We use benchmarks like FrontierMath to measure AI performance, so why not create a dataset to benchmark bias and honesty? It’s not just about capability, people need to trust these systems too.
1
0
2