There are “whole parts of the world” which have been excluded from conversations about how best to regulate and control artificial intelligence, a new report is warning. An international panel of 38 tech experts tasked by the UN Secretary General to advise it on AI issues has today (Thursday) released its findings – and it is highly critical of what it describes as a “patchwork” of existing rules and guidelines (anyone on the AI beat will tell you we have read a *lot* of codes, pledges, commitments on responsible AI in recent years - almost all of them voluntary)   It looked at a number of recent high-profile multi-national AI treaties, including the Bletchley Park declaration and the G20 AI Principles, and found that only Canada, France, Germany, Italy, Japan, the UK and the US had signed up to all of them. More than 100 of the UN’s 193 member states were missing entirely. Many of those are in the global south.   The UN secretary-general’s high-level advisory body on artificial intelligence also criticised the voluntary nature of many existing guidelines which are currently in place in various countries.   “No one currently understands all of AI’s inner workings enough to fully control its outputs or predict its evolution,” the report states. “Nor are decision makers held accountable for developing, deploying or using systems they do not understand.”   It further claimed that “many terms routinely used with respect to AI – fairness, safety, transparency – do not have agreed definitions” leading, it argued, to the creation of industry standards which are “divorced from the values that they were intended to uphold.”   The group,  which included the brilliant UK computer scientist @DameWendyDBE ,  is recommending that the UN should set up its own global AI scientific committee, similar to the Intergovernmental Panel for Climate Change. It suggests establishing an AI office in UN headquarters, and a global AI fund to ensure that more countries are able to access AI tools and training.   “Of the top 100 high-performance computing clusters in the world capable of training large AI models, not one is hosted in a developing country,” it notes.   The UN will now consider the recommendations.   As part of its research, the group also surveyed 340 tech experts and asked them to rate their concerns about the use of AI in 14 different areas. The use of AI to compromise information (such as in the generation of misinformation and deepfakes), and the use of the tech in armed conflict, emerged as the two themes which worried them the most.   There are already growing calls for the UN to update its treaty on autonomous weapons, including from the Red Cross.   “The Advisory Body considers it essential to identify clear red lines delineating unlawful use cases, including relying on AI to select and engage targets autonomously,” the report says.

Good news for team UK AI: Microsoft opening a London hub here blogs.microsoft.com/blog/202…

So cool - I was at Goonhilly Earth Station in Cornwall on Monday which was due to guide the commercial spacecraft down onto the moon ✨so good to see the UK space sector involved in big projects bbc.co.uk/news/science-envir…