visarga

38 ms·

781.
by visarga 2y ago
I don't know about copyright, since for most artists the royalty revenues are not enough to live on. It seems like a failed system if the intent was to get royalty revenues.
782.
by visarga 2y ago
The logic "gen AI is theft" is pretty careless. Let's say I use gen-AI to identify a skin sore, and seek appropriate treatment. Who's copyright was being violated? How about if I ask it to make a story where my kid is a
783.
by visarga 2y ago
> What ML does is not in any way akin to human learning in methodology. When we call it learning that's an analogy. Of course it's different, but if we look closely, it is not copying. The model itself is smaller, sometimes 10
784.
by visarga 2y ago
> By training a model on a copyrighted work, you can launder that work into your own new work, and make money from it without sharing that money with the original artists. Not sure if "you" refers to model developers, hosting c
785.
by visarga 2y ago
> IP is such a stupid concept It's been struggling since the internet became a thing. People got more content than they can consume. For any topic there are 1000 alternative sites, most of them free. Any new work competes against de
786.
by visarga 2y ago
Are you sure you mean "stealing"? As in deprive him of his own recordings? I am curious if anyone read Harry Potter in bootleg form from a LLM. I mean, LLMs are the worst tools for infringing - they are approximate, expensive and
787.
by visarga 2y ago
We move up, down or sideways on the stack. That's the outcome. Not necessarily bad. It requires soul searching to find out new place.
788.
by visarga 2y ago
> There is no such thing as "get as good as a human artist" unless it becomes an actual human that lived the human experience. Even bad art starts with something to express and a want to express it. There is always an actual hu
789.
by visarga 2y ago
> Modern BERT with the extended context has solved natural language web search. I mean it as no exaggeration that _everything_ google does for search is now obsolete. The web UI for people using search may be obsolete, but search is hot,
790.
by visarga 2y ago
I just had very bad JSON mode operation with gemini-1.5-flash and 2.0-flash models using their own library 'google-generativeai'. Either can't follow JSON formatting correctly, or renders string fields with no end until max_t
791.
by visarga 2y ago
Very interesting trick, using a dictionary of basis vectors which are quickly computed from a seed without storage. But the result is the same 3 or 4 bit quantization, with only a slight improvement. Their tiles are small, just 8 or 12 weig
792.
by visarga 2y ago
This is true, a lot of progress can still happen based on simulation and synthetic data. But I am considering the long term game. In the long term we can't substitute simulation to reality. We can't even predict if a 3-body system
793.
by visarga 2y ago
I think you are underestimating the context, we all stand on shoulders of giants. Let's think what would happen if kid Einstein, at the young age of 5, was marooned on an island and recovered 30 years later. Will he have any deep insig
794.
by visarga 2y ago
I have considered this too. I frame it as problem solving. We are solving problems across all fields, from investing, to designing, construction, sales, entertainment, science, medicine, repair. What do you need when you are solving problem
795.
by visarga 2y ago
Yeah I think the math+code reasoning models, like o1 and r1, are doing what can be done with just pure compute without real world validation. But the real world is complex, we can't simulate it. Why do we make particle accelerators, fu
796.
by visarga 2y ago
The story is entertaining, but it has a big fallacy - progress is not a function of compute or model size alone. This kind of mistake is almost magical thinking. What matters most is the training set. During the GPT-3 era there was plenty o
797.
by visarga 2y ago
> I view this as compelling evidence that current models are more than "stochastic parrots," More evidence against "stochastic parrots" - zero shot translation, where LLMs can translate between unseen pairs of languag
798.
by visarga 2y ago
Nobody will want to visit your home if it has trap doors.
799.
by visarga 2y ago
I loved Delphi 1 ages ago but I won't fork thousands of dollars for a language.
800.
by visarga 2y ago
I found that ChatGPT or Claude are really good at music and shopping suggestions. Just chat with them about your tastes for a while, then ask for suggestions. Compared to old recommender systems this method allows much better user guidance.
801.
by visarga 2y ago
Like burning down you house so squatters can't use it. A poisoning attack that makes communication itself untrusted. Wondering if we will get to extreme reciprocal mistrust eventually. This also reminds me of art images doctored to bre
802.
by visarga 2y ago
I don't dispute your claim, just that I still find Alex voice to be the best, and it's been the same since over 10 years ago. The other voices have issues, they don't sound too good at 1.5x.
803.
by visarga 2y ago
Apple's TTS voices still run on 10 year old technology. Pretty disappointing, at one time the had the best system voices.
804.
by visarga 2y ago
I don't care because it doesn get deployed in EU anyway. EU-Siri is just as dumb as it has always been.
805.
by visarga 2y ago
Most interesting data will be collected in chat rooms and apps. There are over a billion LLM users, they act as human-in-the-loop enhancig the LLM, sometimes testing ideas even in reality. Wondering what providers are doing with the chat lo
806.
by visarga 2y ago
Actually even the new model does it. I put it read "12345 54321" and it read "2346 5321". So it both skips and hallucinates digits. This could be dangerous if it is used to read some news article or important text with n
807.
by visarga 2y ago
Hey Jeff, maybe you could improve the TTS that is currently in the OpenAI web and phone apps. When I set it to read numbers in Romanian it slurs digits. This also happens sometimes with regular words as well. I hope you find resources for o
808.
by visarga 2y ago
If you want deeper search, it needs to be able to iterate, plan, reason while searching.
809.
by visarga 2y ago
The cool thing about Conway's Game of Life is that you can't predict it unless you do the full recursion, there is no shortcut. It relates to external undecidability of recurrent processes.
810.
by visarga 2y ago
There are over 200K language modeling datasets on Hugging Face, I bet a large portion of them were generated with LLMs, and all LLMs to date have been trained on copyrighted data. So they are all tainted. But philosophically, I wonder if it

More ›