Greetings,

Considering that there is no standard for test & evaluation of AI models yet, it makes sense that a lot of companies have banned their employees from using generative AI. But I’m sure they’ll be back on the genAI train soon with progress moving as quickly as it is.

• 75% of 2,000 IT decision-makers from 5 continents say their companies have banned ChatGPT and other generative AI tools due to risks associated with data security, privacy, and brand reputation. - ZDNet 
  
  
• There are currently no standards for test & evaluation in AI development – to help develop them Scale, a company partnered with leading AI companies, uncovered their own, detailed approach to T&E. - Scale
  
  
• Check out the demo of Nvidia’s NeVA, a powerful model that understands both text and images. - Nvidia
  
  
• Interesting method to make LLMs more useful for programmers – this library can guide any model, even a tiny one, into generating text that perfectly matches RegEx expressions or a JSON schema. - GitHub, arXiv
  
  
• Authors are finding books they didn’t write sold with their names on them – AI isn’t the main issue here, it’s more about lack of verification and protection for authors, but it does make it much easier to perform fraud like this. - ArsTechnica
  
  
• DoctorGPT is a 3GB model that you can use on iOS, Android, or in the browser for free, offline, without sharing any data – it’s based on Llama 2, and it can pass the US Medical Licensing exam. - GitHub
 
• The future of drive-thrus – see what it’s like to order food in a fully automated drive-thru experience powered by an AI assistant. - Axios
  

----

Found this helpful? Forward this email to a colleague.

Was this email forwarded to you? Sign up here to stay on top of the AI news.