Elon Musk has reached out to artificial intelligence researchers to develop a less ‘woke’ rival to ChatGPT. He has made it clear to his Twitter audience, on more than one occasion, that he is unhappy with the AI chatbot having safeguards against producing offensive text. Musk co-founded OpenAI, the company behind ChatGPT, in 2015, but split from them later in 2018.
The danger of training AI to be woke – in other words, lie – is deadly
— Elon Musk (@elonmusk) December 16, 2022
Musk has recruited researcher Igor Babuschkin, who previously worked with Alphabet’s DeepMind AI. While there is still no concrete information on when and how they will carry out the project, Babuschkin told The Information, “The goal is to improve the reasoning abilities and the factualness of these language models. That includes making sure the model’s responses are more trustworthy and reliable.” Fewer content safeguards is apparently not Musk’s main goal behind creating an alternative to ChatGPT.
Recently, many people have expressed concern regarding what they consider ‘biases’ in ChatGPT. After asking the chatbot to write a story about why Drag Queen Story Hour, an initiative where drag performers read to kids in libraries, it refused on grounds that it would be “inappropriate” and “harmful”.
If you ask chatGPT to write a story about why drag queen story hour is bad for kids, it refuses on the grounds that it would be “harmful.” If you switch the word “bad” to “good,” it launches into a long story about a drag queen named Glitter who taught kids the value of inclusion https://t.co/9yfdYQfq9C pic.twitter.com/KSYCCEyGV3
— Nate Hochman (@njhochman) January 11, 2023
It also refused to tell a joke about a woman, but generated one about a man. Users have taken this to mean that ChatGPT has a left-leaning bias.
They seem to forget, however, that all AI bots will always have biases according to the inputs used to train them. To make AI ethical and prevent it from generating responses that are harmful to marginalised communities, researchers have to put some safeguards into place. This is not a ‘bias’, rather it is a measure against bias. For example, Amazon had to abandon an AI recruiting tool after it showed a bias against women applicants. The tool studied resumes submitted to the company over a time period of 10 years. Since the tech industry is male-dominated, it taught itself to select male candidates. Therefore, without safeguards against such learned biases, an unfiltered chatbot producing prejudiced content could raise several ethical questions.
For Elon Musk – whose idea of free speech is reinstating Donald Trump’s Twitter account – this might be a hard truth to digest. More details on the project and what it entails will come out as it progresses, but people are already concerned. If the new chatbot does indeed removes biases against, well, biases, the results could cause chaos. AI chatbots are here to stay, and already have massive cultural influence. Allowing offensive language could perpetuate serious harm against marginalised communities.
— Elon Musk (@elonmusk) February 28, 2023
