Geoffrey Hinton is known as the Godfather of AI. In the 1980s, he used ideas from Statistical Physics to help create the Boltzmann machine, a neural network that could learn to recognise characteristic patterns in data. In 2024, Hinton, along with John Hopfield, won the 2024 Nobel Prize in Physics for “foundational discoveries and inventions that enable machine learning with artificial neural networks.” Today he is warning about the dangers of advanced AI: misinformation, job disruption, inequality, autonomous weapons and the possibility that machines may eventually escape human control. Yoshua Bengio’s work on representation learning, neural language models and deep learning fed directly into the techniques that now animate translation systems, speech interfaces, recommendation engines, scientific models and generative AI. Today he warns that AI progress is not well managed and not restrained. He has warned that frontier models are beginning to show deception, hacking, self-preservation and goal misalignment. Yann Lecun’s work on neural networks for handwriting and image recognition helped make machine perception practical. He argues that LLMs trained to predict the next token do not understand the physical world in the way animals and humans do. They lack grounded models of reality, persistent memory, planning, causal reasoning and the common sense ability to simulate consequences. They can speak with astonishing fluency, but fluency is not the same as intelligence. The goal of Dario Amodei, Ceo of Anthropic, is ‘a public benefit corporation dedicated to building AI systems that are steerable, interpretable and safe’. His 2024 article “Machines of Loving Grace” tried to describe what might happen if powerful AI works right: faster progress in biology, health, governance, income growth and poverty reduction. Basically, he is asking for disciplined AI and needs to put the brakes on, if needed. Jensen Huang cofounded NVIDIA in 1993 with Chris Malachowsky and Curtis Prem and worked on specialised processors –GPUs were built to process many operations in parallel, ideal for Deep Learning. Ultimately, they built a Complete Ecosystem: CUDA, libraries, developer tools, systems, interconnects, networking and eventually AI factories. Without NVIDIA’s hardware and software stack, much of today’s frontier AI would be slower and more expensive. They have given several warnings on AI covering workforce displacement, energy infrastructure and geopolitical chip restrictions.
Government of the USA is working on developing an AI Safety Order, under which AI companies would inform the Government before releasing new systems. India has also released AI Governance Guidelines and is expected to upgrade the same over time. At the same time, we are getting news that ‘Corporations cut back on AI usage as cost of agentic models strains budgets’ and ‘Companies are Quietly Rehiring Workers the Workers they Fired for AI’. There is a need to strike a balance between AI and people, as AI can support people and not replace it and misuse of AI can be dangerous.
A detailed AI Safety Report is available at https://internationalaisafetyreport.org/.
We have given our views. Now would love to hear from you.