The benefits and opportunities offered by artificial intelligence onewas mentioned director of Computing Systems at the MIT Media Lab, Dr. Michalis Bletsas, commander of the National Cybersecurity Authority, speaking at the Artificial Intelligence conference at the Beyond exhibition. “Five years from now, if we are still only talking about Artificial Intelligence, we will probably be on the wrong path. We should be talking about Extended Intelligence, not about man versus machine, but about man (alongside) the machine,” he said.
As he said, another big question that has opened up is whether we will be able to achieve higher forms of AI, like Artificial General Intelligence, by continuing to develop and expand the scale of large language models (i.e. LLMs, like ChatGPT). . In addition to this issue, he added, there are three major problems. The first is that there is no additional data available (for training AI): “We can generate zeta bytes of data every year, but most of it is now generated by machines, in short, it is not suitable for training AI models, the that they will have to imitate human intelligence,” he explained.
O second problem is that larger computing systems are needed, at a time when “Moore’s Law” has largely… done us in: “We can no longer hope that every year we will have faster computers and processors,” noted Mr. Bletsas and added that many improvements that have been achieved now come from algorithmic advances and this is very important, because this work can still be done within universities, outside of large technology companies. O third problem is that as the IT race accelerates and grows, “we won’t have enough (electrical) power to ‘operate’ the necessary data centers.” It will take a lot of algorithmic and mental advances to find new ways (of developing AI) and I think that the path that the big AI companies are trying to follow today will not take us any further. I think we’re going to hit a wall soon,” he said characteristically.
“Many companies ran”
Regarding the use of AI, especially its so-called “creative”, “generative” or “generative AI” form, Mr. Bletsas expressed the conviction that many companies have rushed to exploit it to a greater extent than what It’s really necessary, especially since people are currently doing a better job. As I said, this zeal is probably due to the fact that (companies) do not want to be left behind, motivated by the FOMO syndrome (Fear of Missing Out) and try to make the most of the Great Linguistic Models (such as ChatGPT), “confusing largely the intelligence itself with its expression”.
TN in “Ágios Savvas”
The professor also recalled that Artificial Intelligence is something very old – its founding conference took place in 1956 – that has been saving lives for years, for example, with its use in cancer diagnostic tests and in radiology. In fact, to reinforce TN’s possibilities in this field, Mr. Bletsas showed the audience two mammograms three years apart. In the first case, the cancerous tumor was already visible, but the radiologist was unable to distinguish it because it was very small and not easily visible. So a professor, wondering why the radiologist didn’t see the tumor, went to Massachusetts General Hospital, which has a highly developed electronic health record system, and collected data from several tens of thousands of mammograms and their results. He then trained a statistical machine learning model to estimate a test subject’s risk of developing the disease over a five-year period. By using models like this, “on the one hand we ‘catch’ possible cancers and on the other hand we reduce the burden on the healthcare system, because unnecessary tests are not carried out (…) A very useful application of simple AI that does not need resources We have validated this model with data from many other hospitals around the world (…) So far auxiliary models have captured 40%-44% of cases, this one captures 80%. Furthermore, the model is open source and MIT gives it free of charge to anyone who wants to use it”, he noted and recalled that together with a Greek company, which produces the diagnostic image management system, which is used in two thirds of NHS hospitals, we are making an effort to validate the model with Greek data at “Agios Savvas” and then distribute it free of charge to all other NHS hospitals”.
Bletsas added that the AI application that has “intrigued” him the most lately has to do with proteins and Deep Mind’s AlphaFold 2 protein algorithm, which are known to play a key role in the functioning of the cells of living organisms. began in 2016 and in July 2022 “AlphaFold” protein structure predictions were made public for almost every recorded protein known to science, largely solving one of biology’s greatest mysteries and opening new avenues for researchers and laboratories around of the world). “Due to this progress, we can already see how much publications in the relevant field of research have increased,” he said, adding that IT applications are also important in cybersecurity, where they can be used, among other things, for educational scenarios and training exercises. training and actions that cannot be processed easily and quickly by humans.
Source: AMPE