Isn’t this relatively trivial to correct? Just like chain of thought reasoning replaces end tokens with “hmm” to continue the thought can’t users just replace the llm tokens whenever it starts saying “maybe they are referring to” with something like. “Let me ask a clarifying question before I proceed.”
At least one hand is operating a circular saw and there’s often a spring loaded guard that snaps into place when the saw isn’t sawing. I’m sure plenty of people hurt themselves with circular saws, but as more of a hands-on tool you are very aware when using it. It is not nearly as precise or easy to use with large jobs as a table saw, however.
On a circular saw, the hand in motion maneuvers the cutting tool. On a table saw, it pushes the thing being cut. Either saw's blade sometimes grabs the material and kicks it back. If your hand is on that material, it can also be sent in an unexpected direction.
(This is just one scenario. Both tools are capable of unwanted removal of body parts.)
Can you clarify what you mean by using an LLM to “get a list that would even out the numbers”? If you’re doing binary classification, you need datapoints for features as well as the target class so how does an LLM synthetically create that without causing problems?
It’s kinda easy to check it though. Ask your friends if they use it, preferably non developer. When I find out my accounting friends and marketing friends are using it for basic analytics/promotion write ups I realized this is being used by non developers already on a regular basis. The use by non tech folks is significant for this . Crypto as a counter example of hype hasn’t lived up to its purported use case as non tech ppl I know only “own” it through an exchange.
I did. Most of them if not all have tried it once just out of curiosity. None of them is using it on a regular basis- not even once a week. Maybe I am in the wrong group? Includes family / friends (non tech). Tech people (friends) use it regularly for coding though.
Basically no one uses FP32 at inference time. BF16/FP16 is typically considered unquantized, whereas FP8 is lightly quantized. That being said there's pretty minimal quality loss at FP8 compared to 16-bit typically; Llama 3.1 405b, for example, only benchmarks around ~1% worse when run at FP8: https://blog.vllm.ai/2024/07/23/llama31.html
Every major inference provider other than Hyperbolic Labs runs Llama 3.1 405b at FP8, FWIW (e.g. Together, Fireworks, Lepton), so to compare against FP32 is misleading to say the least. Even Hyperbolic runs it at BF16.
With serious diminishing returns. At inference time, no reason to use fp64 and should probably use fp8 or less. The accuracy loss is far less than you'd expect. AFAIK Llama 3.2 3B fp4 will outperform Llama 3.2 1B at fp32 in accuracy and speed, despite 8x precision.
I believe Anthropic was the most acclaimed/highest current value, and they still have about 1/3 of the original 8% investment.
a) I am curious at what discount they sell Anthropic shares right now;
b) I am not sure what is the prognosis for the other shares FTX owns;
[
c) which creditors have preference to which money pot; usually there is a pyramid. Here it talks about consumers. (Specifically those owed less than $50k.)
And there is a separate matter for shareholders who are looking at getting some of the seized by DoJ proceeds. (https://www.reuters.com/legal/crypto-exchange-ftxs-liquidati....)
]
Best way in general is to read the actual pleadings: news is not the best at giving a good idea what the court is actually ordering.
