-
Asked by anon-379276 on 18 Jan 2024.
-
Alexander Coles answered on 18 Jan 2024:
It depends on what you mean by “beyond human capacity”. AI’s huge advantage is its ability to efficiently learn to do extremely specific tasks from loads of data that may take a human years to go through. This has meant that in some cases for example in the radiology of breast cancer tumours, an AI model has been able to identify cancers from an image that a radiologist might miss, but then again the same AI model has also missed tumours that are glaringly obvious to a trained radiologist. We do hope that we can equal or surpass the ability of a human in these life-saving fields however it will always be wise to have a trained human specialist in the loop to spot something an AI might miss.
-
-
Luke Humphrey answered on 18 Jan 2024: last edited 18 Jan 2024 11:14 am
One potential AI application which is beyond human capacity is in plasma control. In a fusion power plant, energy is produced by fusion reactions in a superheated plasma. No materials can contain the plasma without melting, instead we use magnetic fields. The plasma “wants” to expand and cool down and become an inert gas, we want it to remain a plasma so fusion reactions occur and release high energy neutrons. In order to maintain control, adjustments need to be made the strength of our magnets based on what the plasma is doing. This is similar to balancing a ruler on your fingers and having to adjust your hand. Humans are limited by their reaction speed and knowledge, while an AI tool trained to make the correct adjustments based on inputs from diagnostic sensors could enable us to maintain a fusion plasma for long periods of time without needing to restart, allowing a fusion power plant to produce lots of energy consistently.
I will add that this is no risk to human life. It happens all the times in our research devices. But it damages the walls of the device, which means expensive repairs during which time the device can’t operate. For a commercial fusion power plant putting electricity on the grid, the goal would be for this to never happen at all. This is important to keep the cost of electricity competitive with fossil fuels so that we can make the case to build carbon free power plants instead of fossil fuel power plants. AI tools will play a crucial role in achieving that goal in my opinion.
This is something a lot of my colleagues are actively researching.
-
Sheridan Williams answered on 18 Jan 2024:
Interesting – will AI ever be able to “think”?
I’m not sure, maybe you need to define “think”. Thinking implies consciousness which I doubt present computers with AI will ever be capable of. Maybe quantum computers, or more likely quantum computers’ successors; ones made of carbon neural networks. -
Carl Peter Robinson answered on 18 Jan 2024:
What an interesting question. In its current form, I wouldn’t say that AI can think, but there are examples of several AI models that have been used to discover things that it would have taken humans years and years to find out. An example is Google DeepMind’s AlphaFold, which predicts different protein structures. It has performed so well that it could help in the creation of new medicines (https://www.theguardian.com/technology/2022/jul/28/deepmind-uncovers-structure-of-200m-proteins-in-scientific-leap-forward). So, while AI cannot think, it can help we humans to discover new solutions that we might never have thought about before, much more quickly, to solve some of the many difficult problems in the world.
As for AI in the future. Who knows? I guess it depends on what happens when we finally crack artificial general intelligence (AGI). That’s certainly something to keep an eye on if you’re interested 🙂
Comments