This week, former employees of OpenAI voiced concerns to The New York Times about safety issues within the company. They say that OpenAI isn’t doing enough to prevent its AI systems from becoming dangerous, and they claim the company is trying to silence whistleblowers.
The ex-staffers published an open letter calling for AI companies, including OpenAI, to be more transparent and offer more protection for whistleblowers. But I’m not very optimistic that their voices will be heard.
OpenAI started a safety commission after facing initial criticism but it’s all made up of people who work at the company—like CEO Sam Altman, who said he didn’t know about the restrictive agreements at OpenAI but actually signed them.
On to other news about AI this week:
- ChatGPT and other AI chatbot platforms experienced an outage but were restored shortly after.
- OpenAI is exploring a deal with fusion startup Helion Energy to buy electricity for its data centers.
- The cost of data licensing for AI is becoming too expensive for smaller organizations and academic institutions.
- Malicious actors are using AI-powered music generators to create offensive songs.
- AI startup Cohere raised $450 million from big names like Nvidia and Salesforce Ventures.
Research of the week highlights how OpenAI is training its AI model, GPT-4, to solve math problems, while the model of the week is Microsoft’s Aurora, an AI model that’s helping improve weather predictions.
Finally, in a thought-provoking opinion piece, a chief of staff at AI startup Anthropic suggests that knowledge workers may no longer be needed in three years due to rapid advances in generative AI. But I’m not so sure I share that optimistic view based on the potential impact on the workforce.
This week has been a rollercoaster of news in the AI world, with some serious concerns being raised about safety, while advancements in AI models continue to make strides. We’ll continue to keep an eye on these developments in the weeks to come.