In other AI news, Microsoft found flaws in OpenAI’s GPT-4, OpenAI launched its internet-browsing feature to ChatGPT and transitioned DALL-E 3 into beta, and two open source alternatives to GPT-4V were released. Google introduced an AI-powered language tutor, Amazon announced plans to test Agility’s bipedal robot, Digit, in its facilities, and China-based startup Zhipu AI raised $340 million in financing. Additionally, AI was used to reconstruct a scroll destroyed in Pompeii, suggest citations on Wikipedia, solve mathematical problems, and decode brain activity.
Key takeaways:
- Google-owned AI R&D lab, DeepMind, has released a paper proposing a framework for evaluating the societal and ethical risks of AI systems. This comes ahead of the AI Safety Summit, a U.K.-government-sponsored event focusing on managing risks from advances in AI.
- DeepMind's parent company, Google, scored poorly in a recent study by Stanford researchers that ranked AI models on transparency. Google's text-analyzing AI model, PaLM 2, scored only 40% on 100 criteria, including disclosure of training data sources and hardware information.
- DeepMind is taking steps to change the perception of its lack of transparency by committing to provide the U.K. government with "early or priority access" to its AI models to support research into evaluation and safety.
- Other AI news includes a Microsoft study finding flaws in OpenAI's GPT-4, OpenAI launching its internet-browsing feature to ChatGPT, and Google introducing a new feature to help people practice their English speaking skills. Additionally, Amazon announced it will begin testing Agility's bipedal robot, Digit, in its facilities.