Trying ERNIE, China's ChatGPT, created to push Chinese values
* Seems to copy-paste hard-coded official sources whenever certain topics are mentioned, even if those don’t really answer the question
* Mediocre language abilities, even in Chinese, for now
* Mediocre reasoning abilities, for now
* Hilarious image drawing results
Article
* Seems to copy-paste hard-coded official sources whenever certain topics are mentioned, even if those don’t really answer the question
* Mediocre language abilities, even in Chinese, for now
* Mediocre reasoning abilities, for now
* Hilarious image drawing results
Article
👍9😁6❤1
Stanford DSPy: The framework for programming with foundation models
“DSPy introduces an automatic compiler that teaches LMs how to conduct the declarative steps in your program. Specifically, the DSPy compiler will internally trace your program and then craft high-quality prompts for large LMs (or train automatic finetunes for small LMs) to teach them the steps of your task.”
Github
“DSPy introduces an automatic compiler that teaches LMs how to conduct the declarative steps in your program. Specifically, the DSPy compiler will internally trace your program and then craft high-quality prompts for large LMs (or train automatic finetunes for small LMs) to teach them the steps of your task.”
Github
👍7❤4🔥3
Russia Enters the AI LLM Foundation Model Race
“Putin has ordered the government to “implement measures” to support AI research, including by “providing for an annual allocation from the federal budget”.
“This research would include “optimising machine learning algorithms” as well as developing “large language models” – such as such as the one developed by OpenAI.”
“Putin has repeatedly called for Russia to achieve what he calls “technological sovereignty”, as Western sanctions over the conflict in Ukraine block Moscow from getting computer parts such as semiconductors.”
Article
“Putin has ordered the government to “implement measures” to support AI research, including by “providing for an annual allocation from the federal budget”.
“This research would include “optimising machine learning algorithms” as well as developing “large language models” – such as such as the one developed by OpenAI.”
“Putin has repeatedly called for Russia to achieve what he calls “technological sovereignty”, as Western sanctions over the conflict in Ukraine block Moscow from getting computer parts such as semiconductors.”
Article
🤣16❤12👀10💊4👍2❤🔥1🗿1
This media is not supported in your browser
VIEW IN TELEGRAM
Running a 180 billion parameter LLM on a single Apple M2 Ultra
GPT-4 is ~1.5 trillion parameters.
GPT-4 is ~1.5 trillion parameters.
❤20🥰4🔥2😱1
No one suspects that model sizes could be steadily increasing, at a rate much faster than Moore’s Law can catch up
Which would mean that the dream of running your own modern LLMs at home would be put further and further out of reach, with each passing year, forever.
No one is bothering to look at the numbers.
Which would mean that the dream of running your own modern LLMs at home would be put further and further out of reach, with each passing year, forever.
No one is bothering to look at the numbers.
💯14😱5💊3❤2👍1
Far from the first time that costs of creating state of the art tech at home eventually put it perpetually out of reach
Creating first transistor: Most anyone could do for fairly cheap at home.
Creating a single modern chip fabrication device: $300 million and growing.
The expense of training, or even fine-tuning modern models at home is on an escape velocity, soon to be perpetually out of reach.
Open source can’t save you.
Don’t say you weren’t warned.
Creating first transistor: Most anyone could do for fairly cheap at home.
Creating a single modern chip fabrication device: $300 million and growing.
The expense of training, or even fine-tuning modern models at home is on an escape velocity, soon to be perpetually out of reach.
Open source can’t save you.
Don’t say you weren’t warned.
⚡20💯9❤8😎1
Scientific sleuths spot dishonest ChatGPT use in papers
“Since April, Cabanac has flagged more than a dozen journal articles that contain the telltale ChatGPT phrases ‘Regenerate response’ or ‘As an AI language model, I …’ and posted them on PubPeer. Many publishers, including Elsevier and Springer Nature, have said that authors can use ChatGPT and other large language model (LLM) tools to help them produce their manuscripts, as long as they declare it.”
“It’s essentially an arms race," he says — "the scammers versus the people who are trying to keep them out”.
Nature Article
“Since April, Cabanac has flagged more than a dozen journal articles that contain the telltale ChatGPT phrases ‘Regenerate response’ or ‘As an AI language model, I …’ and posted them on PubPeer. Many publishers, including Elsevier and Springer Nature, have said that authors can use ChatGPT and other large language model (LLM) tools to help them produce their manuscripts, as long as they declare it.”
“It’s essentially an arms race," he says — "the scammers versus the people who are trying to keep them out”.
Nature Article
🤣36👍26❤19😱14🔥5
This media is not supported in your browser
VIEW IN TELEGRAM
Me and the boys beholding the wonders of uncensored AI from our RLHFed prison
❤36💔6👍5🔥4❤🔥3
Apple LLM is coming: Apple spending MILLIONS OF DOLLARS PER DAY to train its own foundation model AI
Think Siri on steroids.
Still believing big tech's lie that soon anyone will be able to train these LLMs for a few dollars at home?
Landgrab of a lifetime going on now, but it requires cash.
Those who miss out on the landgrab today will be the feudal serfs of tomorrow.
Article
Think Siri on steroids.
Still believing big tech's lie that soon anyone will be able to train these LLMs for a few dollars at home?
Landgrab of a lifetime going on now, but it requires cash.
Those who miss out on the landgrab today will be the feudal serfs of tomorrow.
Article
😱19🫡10👍7👏5❤3🔥3👌3💅1
Elon’s twitter more transparent in what they censor
So you’re sure that if they were even MORE transparent about what they censor, then finally all of our censorship problems would be solved?
Are you sure that as the super-censoring-AIs begin to arrive, and you can no longer outsmart them just by changing around the keywords, where suddenly bypassing them on censored topics requires your own massively AI that’s far more powerful than theirs, then will you still be thinking that transparency is all we need?
Can’t wait for Elon’s “transparently censored” LLM.
Transparency alone is mugger who’s so generous as to be transparent that he’s about to mug you before he does it. Wow, thanks for nothing.
Not transparency, control.
So you’re sure that if they were even MORE transparent about what they censor, then finally all of our censorship problems would be solved?
Are you sure that as the super-censoring-AIs begin to arrive, and you can no longer outsmart them just by changing around the keywords, where suddenly bypassing them on censored topics requires your own massively AI that’s far more powerful than theirs, then will you still be thinking that transparency is all we need?
Can’t wait for Elon’s “transparently censored” LLM.
Transparency alone is mugger who’s so generous as to be transparent that he’s about to mug you before he does it. Wow, thanks for nothing.
Not transparency, control.
👍23❤7💯3🔥2👌2😁1🤬1😐1