Who’s afraid of ChatGPT? The competition, apparently.
Over 1,000 academics and techies signed an open letter calling for companies to pause training on any training of AI models “more powerful” than GPT-4 for six months. While the letter is getting a lot of attention, what’s overlooked is how rapidly shrinking tech companies have gutted their ethical AI departments in the face of ChatGPT’s growing “power,” per a report in the Financial Times.
Ethics 101 — Don’t Fire The Ethicists
The open letter garnered signatures from notable tech leaders including Elon Musk, researchers from Google-owned DeepMind, and the CEO of Stability AI — the company that made AI-powered image generator Stable Diffusion. Notably, these signatories are all in direct competition with OpenAI.
Musk disbanded the ethical AI team at Twitter when he fired the vast majority of the company’s workforce — and he’s not alone. Google, Meta, Amazon, and Microsoft have all made cuts to their AI ethics teams, according to the FT. While companies may be shedding AI ethics voices internally, some experts have given their opinions on the open letter publicly:
- Joanna Bryson, an AI expert who publicly criticized Google after it disbanded an ethics council she sat on, tweeted about the open letter: “We don’t need AI to be arbitrarily slowed, we need AI products to be safe. That involves following and documenting good practice, which requires regulation and audits.”
- AI policy expert Dr. Nakeema Stefflbauer told The Daily Upside she found the letter disingenuous. “The line ‘more powerful than GPT-4’ really gives away the game,” she said. “Fact is, GPT-4 has already paved the way for rampant disinformation and IP devaluation, so an effective moratorium on AI development would require a rolling back of that technology, as well.”
Stefflbauer pointed out moratoriums on extant AI tech are not a new phenomenon. In 2020, major US tech firms paused sales of facial recognition to US law enforcement following activist pressure from AI ethics experts who pointed out bias, particularly the tendency to misidentify people of color and women.
Move Fast And AI-ify Things: The chances of Microsoft and OpenAI slowing down are pretty slim. Microsoft seems keen to cram ChatGPT into as many products as possible, and this week announced it will be integrated into its cybersecurity suite of products, though it was cagey about exactly what the chatbot’s use will be. Maybe ChatGPT can invent one.