Maybe OpenAI Should be Called ClosedAI

ChatGPT 3.5 was basically a beta. They got lots of people to feed data into the system, including sensitive, proprietary data (oops) in order to train it.

Now that it is “more” trained, OpenAI released ChatGPT 4.

Contrary to the founding principles of investors like Elon Musk, the company, which claimed that it would be dedicated to research that advances humanity as a whole, unconstrained by a need to generate a financial return, is now like any other business.

Microsoft invested $10 billion into OpenAI and there is an AI war going on.

OpenAI released a 98 page document on Tuesday along with the unveiling of its new product. But, unlike the beta, there really is no useful information in the document according to people who have read it.

The report actually says  “Given both the competitive landscape and the safety implications of large-scale models like GPT-4, this report contains no further details about the architecture (including model size), hardware, training compute, dataset construction, training method.”

Of course, some information will come out and competitors will reverse engineer it, but, with a large language model, so much is dependent on the data used to train it. That won’t come out unless they get hacked.

One researcher said this about the change in their business model:

Ben Schmidt, the Vice President of Information Design at Nomic, tweeted that OpenAI’s failure to share its datasets means it’s impossible to evaluate whether the training sets have specific biases. “To ameliorate those harms, and to make informed decisions about where a model should *not* be used, we need to know what kinds of biases are built in. OpenAI’s choices make this impossible,” . 

https://www.vice.com/en/article/ak3w5a/openais-gpt-4-is-closed-source-and-shrouded-in-secrecy

Another researcher said this:

“After reading the almost 100-page report and system card about GPT-4, I have more questions than answers. And as a scientist, it’s hard for me to rely upon results that I can’t verify or replicate,” Sasha Luccioni, a Research Scientist at Hugging Face, told Motherboard. Hugging Face is a company that provides open-source tools for building applications and training sets for machine learning. 

https://www.vice.com/en/article/ak3w5a/openais-gpt-4-is-closed-source-and-shrouded-in-secrecy

OpenAI’s chief scientist has been trying to spin all this. He said that it is competitive out there. GPT-4 is not easy to develop. It took pretty much all of OpenAI working together for a very long time to produce this thing.

But then he went on to say “we were wrong” to have called themselves open-source. “I fully expect that in a few years it’s going to be completely obvious to everyone that open-sourcing AI is just not wise”.

I am sure this is not the end of this conversation and the article has more information, but I think the complaint is that it seems that they are changing the game. No one would have complained if OpenAI called themselves ClosedAI at the beginning, no one would be blinking an eye.

ChatGPT and Bing Chat are having problems. Remember that Google couldn’t even get their demo to work. The technology is very, very young, so this should not be a big surprise.

Keeping the dataset secret does have its advantages. People won’t be able to tell, as easily, if OpenAI stole their data to train the software. The upside of this is fewer lawsuits against the company, which they probably think is a plus.

Right now, OpenAI plans to charge between $20 and $40 a month per user. I am sure people will share accounts, but the problem with that is that it will muddy the results. If the software is trying to learn and it has multiple people throwing questions at it and it thinks that they are all the same person, well, that likely has problems.

Stay tuned. This is going to be a very competitive business, so who knows what tomorrow will bring. Credit: Motherboard

by