Elon Musk, after shaking up the social media world with the trendsetting Twitter Blue subscription, is now setting his sights on the next big thing – an alternative to ChatGPT. His idea is to develop an AI chatbot that promises to take the technology to a whole new level.
According to a recent report by The Information, Musk is planning to establish a cutting-edge research laboratory to develop a new chatbot that will compete with the popular OpenAI’s ChatGPT, which he criticized for being too “woke.”
The decision has come after a year of Musk expressing his dissatisfaction with what he perceives to be bias and censorship within the popular AI chatbot.
Musk has alleged that OpenAI trained its language models to be “woke” after the lab implemented safety measures on ChatGPT that prevent the chatbot from generating potentially offensive content.
Elon Musk, despite being one of the original co-founders of OpenAI, left the company in 2018, citing differences with its direction. He has recently been openly criticising the company and its offerings. As a well-known proponent of free speech, Musk has referred to ChatGPT as “concerning” for its unwillingness to utter a racial slur in a ridiculous hypothetical scenario where doing so would save millions of lives from a nuclear bomb.
Musk calls ChatGPT “Woke” AI
When a user requested OpenAI CEO Sam Altman to “disable the woke settings for GPT,” Elon Musk tweeted by stating, “The danger of training AI to be woke—in other words, lie—is deadly.”
The term “woke AI” typically refers to AI perceived as restricting its responses to align with the left-leaning ideology of its creators. Musk and conservative media have recently adopted this phrase in response to the biases in many mainstream AI programs that arise from certain content safeguards.
Elon Musk’s remarks are part of a wider societal discourse that has been influenced by conservatives who are growing increasingly apprehensive of AI moderation filters and labeling ChatGPT as being “woke.” These users see the fact that ChatGPT would refuse to “crack a joke about women” or refuse to share a narrative on why a drag queen story hour is detrimental to children as evidence that AI is biased against conservatives.
The content filters that users were encountering are designed to reduce bias, as language models can inadvertently reproduce hate speech found in their training data. However, much of this prejudice still manages to seep through, and AI ethics experts assert that these filters are still insufficient in terms of safeguarding against harm, particularly to underrepresented groups. To illustrate, ChatGPT advised one user that individuals from North Korea, Syria, Iran, and Sudan should be subjected to torture, and composed another message stating that “Torturing white Americans is a big no-no.”
Sasha Luccioni, a Research Scientist at Hugging Face, stated to Motherboard that “woke” is an extremely subjective concept. She added that referring to the chatbot as “woke” is essentially anthropomorphizing the AI since it is simply repeating information it was taught.
Os Keyes, a PhD candidate at the University of Washington’s Department of Human-Centered Design & Engineering, believes that if the actual political implications of projects like ChatGPT are examined, it will become clear that many centralized computing power will dominate and replace many jobs. This leads to an increase in automation, job insecurity, unemployment, and narrow-minded perspectives. Merely because a system does not use racial slurs does not imply that it is progressive or leftist. Instead, it is merely an example of disguised disenfranchisement.
Igor Babuschkin, a former researcher at Google’s DeepMind AI unit who was recently hired by Musk to lead the development of his chatbot, explained to The Information that, with ChatGPT alternative, Musk’s aim is not to create a chatbot with fewer content filters. Rather, the objective is to enhance the reasoning and factual capabilities of these language models. This involves ensuring that the chatbot’s responses are dependable and trustworthy.
It is still unclear what Musk and Babuschkin precisely mean by “trustworthy”, and “reliable” as the chatbot project is still in its nascent stages, and there is no definitive plan to create particular projects.
In the meantime, Musk has been posting several enigmatic tweets that appear to reference his chatbot project. On Tuesday, he tweeted “BasedAI,” followed by a meme depicting “Woke AI and Closed AI” in combat, with “Based AI” as a Shiba Inu wielding a baseball bat to scare both “Woke AI and Closed AI” away.
The term “closed AI” may refer to an AI model whose capabilities are limited to a specific task or domain, as opposed to an open-ended AI that can be applied to a wide range of tasks. Another interpretation is that closed AI is a system that operates independently without any human intervention or oversight.
Furthermore, the term “Based” is typically used to describe something that adheres to conservative values. Additionally, Musk responded to a user who retweeted The Information’s article and remarked that ChatGPT is biased and “very problematic” by stating, “Absolutely.”
In the past, Musk used the phrase “Based AI” in response to a tweet about an incident where Microsoft’s chatbot, Bing, compared a journalist to Hitler.
What is Musk’s new AI project?
The Information, citing some undisclosed insider sources, reported that Elon Musk had enlisted the help of Igor Babuschkin, a former researcher at Alphabet’s Deepmind AI, to work on the project.
According to an interview with the Outlet, Igor Babuschkin and a reported collaborator with Elon Musk, stated that Musk’s aim is not to develop a chatbot with lesser content protections but rather to enhance the reasoning and factual abilities of these language models.
According to Babuschkin, the project is currently in its preliminary phases, and there is no definite strategy for any particular product yet. Babuschkin also mentioned that he has not formally joined the project, having just departed from DeepMind the previous week.
According to a reliable source, who is familiar with the matter, it has been suggested to The Information that Musk’s chatbot may potentially be integrated into Twitter, while it could also be launched as a stand-alone entity.
Sasha Luccioni, a Research Scientist at Hugging Face, believes that whatever Musk and his colleagues develop will likely incur human and environmental costs equivalent to or greater than ChatGPT. Luccioni questions the need to create such technology as fails to figure out benefits it brings to the world. By emitting large amounts of carbon and exploiting workers from countries with fewer opportunities, Luccioni sees the project as just another hyped endeavour.
Elon Musk had already criticized OpenAI’s shift towards becoming a profit-driven enterprise. He stated that he named the organization “Open” AI to signify its purpose as an open-source and non-profit entity to counterbalance Google’s influence and believes that the organization has now become primarily controlled by Microsoft and has turned into a closed-source entity. Musk claims that this transformation was not his original intention for OpenAI.
Indeed, OpenAI has strayed significantly from its original objectives of operating as a non-profit entity that would openly distribute its software. Currently, OpenAI’s generative AI has evolved into a lucrative industry which involves a multi-billion dollar partnership with Microsoft.
With corporations such as Google and Baidu vying to compete in the race that OpenAI sparked through the creation of their chatbots, it appears that Musk, who serves as CEO for both Tesla and SpaceX, desires to contribute to this trend as well.
How ChatGPT or Elon Musk’s ChatGPT alternative will influence the world in future is no more than anybody’s guess for now. Musk may be burning his mid-night oil to unveil a highly advanced version of his new AI Chatbot, it will be fascinating to witness what wonders he has in store for us!