The most remarkable feat achieved by Sam Altman, 37, in the chatbot and its successor chatbot HTML4 was almost unnoticed.
In 2015, he convinced Elon Musk that he would donate $100 million to OpenAI, an AI non-profit organization. The charity was then transformed into a business, which attracted a $10B investment from Microsoft. He also abandoned the original principle of OpenAI: that it would “open source”, that is, share its technology, for “the good and humanity” and not be “constrained by the need to generate a financial return.
Altman instead has blocked developer access, launched multiple products, amassed more than 100,000,000 users, and started charging $20 per month for enhanced OpenAI chatbot access. Musk reacted last week, saying: “I am still confused as how a non profit to which I donated $100M somehow became $30B market cap. Why isn’t everyone doing it?
GPT-4 was released last week, which is a chatbot that has stunned the world. This release demonstrates OpenAI’s remarkable conversion. It was accepted by nearly every university and advanced-study entrance exam. It turned a photo of handwritten instructions into a functional website and convinced a gig worker it was someone with a vision impairment who needed assistance logging onto websites.
The San Francisco company published a 99-page report on GPT-4’s capabilities as part of the splashy release. It also excluded any data that could help outsiders understand its workings or identify its blind spots. GPT-4 is, in other words a black box.
“Both the competitive landscape as well as the safety implications of large scale models like GPT-4,” the company stated. It did not provide any details about the architecture (including model sizes, hardware, training computes, dataset construction, training methods, or similar”.
OpenAI’s argument was not accepted by everyone. OpenAI’s cautious approach to artificial intelligence has led to fierce debate about how this technology — which insiders claim will soon become as widespread as electricity and be at least as transformative — should develop.
Emad Mostaque is an ex-hedge fund manager and London-based Stability AI developer of Stable Diffusion. He has been Altman’s most vocal critic after Musk.
Mostaque’s views can be summarized to these two points: AI is the greatest technology ever made, and it is infrastructure that will allow new industries and capabilities to rise. It must not be monopolized by one company, but open-sourced.
The 39-year old trolled OpenAI last week amid all the GPT-4 hullabaloo with a tweet attempting to lure his San Francisco rival’s talent. He wrote, “Open offer to any OpenAI employee who wants to work on OpenAI.” We will match your salary and benefits, but you can also work on any other open-source AI projects, whether they are ours or theirs. Collaboration, openness and prioritisation of good outcomes over personal gain are key principles.
OpenAI claims that open-sourcing its technology does not make economic sense. According to Ilya Sutskever, its chief scientist, OpenAI’s founding promise to share its technology was “wrong”. Why? OpenAI believes that we are closer to artificial general Intelligence (AGI), a super-intelligence that is just as good or better than humans at cognitive tasks, than most people realize.
Sutskever stated that “in a few years it will be completely obvious to everybody that open-sourcingAI is just not wise.” It will become quite simple, if one wants, to do a lot of damage with these models. It makes sense to not disclose these capabilities as they get more powerful.
OpenAI’s position boils down to this: “Trust me.” This technology is frightening, and we will ensure it is not misused or falls into the wrong hands.
Mostaque is the exact opposite. The majority of the technology that underpins the internet is open source. This means it can be freely modified, improved, and improved as the needs change. It is now possible to access a robust and affordable technology from any smartphone. It’s a crowd-sourced resource.
Mostaque believes that AI should be identical. OpenAI is not allowing experts to look beyond its walls. How can we possibly know which pitfalls are lurking?
We will have to trust Altman who conned Elon Musk into turning a charity into $30 billion.