Business
MIT Technology Review recently released its first-ever AI-only list‚ '10 things that matter in AI right now'.
Soumitra Dutta‚ former dean of Oxford's Said Business School and IIT Delhi alumnus, has followed the developments the list features․ He has spent decades examining how technology changes economies‚ institutions‚ and societies‚ so can distinguish signal from noise․
The signal Dutta finds most compelling‚ one that most organizations are still treating as a future concern‚ is agent orchestration‚ the coordination of multiple AI systems working together autonomously towards complex‚ long-horizon goals․ The era of the single chatbot‚ useful as it has been‚ is giving way to something structurally different: AI systems that plan‚ delegate‚ and adapt across tasks without waiting for human instruction at each step․ "This is what will reshape industries -- not a better autocomplete‚ but systems that can plan‚ delegate‚ execute‚ and adapt across long-horizon tasks," says Dutta, who holds a PhD in computer science from the University of California, Berkeley.
A trend on the list Soumitra Dutta says is underappreciated is China's decision to pursue open-source AI and give free frontier models to the global developer community. He calls it a masterstroke․ Once any country's infrastructure is substantially based on this particular kind of AI‚ it's going to be very hard to regulate out of it․ The geopolitical implications have not been widely appreciated in most boardrooms outside of Asia‚ in his view․ The window for the West to reckon seriously with this is narrowing․
In research, Dutta sees a compression of time which should compel conversations. AI systems that run experiments‚ synthesize bodies of scientific literature‚ and iterate hypotheses are moving from laboratory curiosity to early deployment․ In fields that will be among the first to be disrupted - such as drug discovery‚ material science‚ climate tech - leaders "need to be asking not 'will AI change our R&D process' but 'how do we restructure our teams to work alongside AI co-scientists effectively?'".
Then there is a risk that Soumitra Dutta singles out as “dangerously” underestimated. Deepfakes - produced at scale‚ deployed in propaganda‚ and used to produce non-consensual imagery - have shifted from the domain of theory to something with real-world effects․ Their growth is outpacing institutional‚ legal and social defenses․ "This is not a technology problem. It is a governance emergency," says Dutta, co-creator of the Global Innovation Index and the Network Readiness Index.