Your weekly update on how AI is changing our lives. Our experts keep it clear and simple, so you can stay ahead of the game. This week we are focussing on medicine. Don't forget to explore our Archive and Share & Subscribe with your friends!
Zero-shot through the heart đ đŻ
Big news has arrived from Europe, with a fascinating study by Swedish radiologists into the benefits of using AI assistants in cancer screening. The study compared the standard practice of 'double reading' - where two radiologists independently assess a patient, before comparing notes - with an AI-assisted screening process which only required a second human opinion in complex cases. They found that the AI-driven success rate was almost identical to that achieved with the traditional approach, but crucially, the work was carried out twice as fast, and without any increase in misdiagnosis.
This is exciting news, but thanks to the pace that AI moves at, itâs also in danger of becoming old news. The study covers a period from April 2021 to July 2022, and a lot has happened since then; just this week Google announced a new AI model with a paper titled 'Towards Generalist Biomedical AI'. The new AI - Med-PaLM M - may sound similar to Med-PaLM, launched just a few months ago, but the newly acquired M in its name is a big deal. If youâve been following chatbot trends closely you may already have guessed what it stands for: 'multimodal'.
Even if youâre not familiar with the term, youâre likely to have encountered multimodality in chatbots already; it simply means the ability to handle many types of data in prompts, rather than just plain text. In Med-Palm Mâs case this is an enormous upgrade because medical data comes in many forms; while the Swedish radiologists in the distant past of 2022 were achieving remarkable results with dedicated AI screening tools, Googleâs new model is far more flexible, and equally comfortable reading reports, viewing X-rays, crunching genomic data and more.
Hereâs where it gets weird. This flexible approach enables whatâs called 'zero-shot reasoning', where the AI is able to identify medical issues that are not part of its training. In Googleâs study, the issues being spotted were signs of tuberculosis in chest X-rays, and Med-PaLM M almost matched the dedicated TB-spotter AI accuracy rate, despite having never been directly trained on the subject. This promises huge efficiency benefits across the entire healthcare sector, and should help to ease punishing workloads while improving patient outcomes. Google is at pains to stress that while certainly impressive, results will need to improve before the system is ready for general use, but the direction of travel doesnât just seem to be 'Towards Generalist Biomedical AI'; we also seem to be accelerating.
Of course, itâs not just medicine where AI is revolutionizing science; donât miss our in-depth look at the various ways that machine learning is boosting the speed of scientific discovery!
Beta Testers Wanted!!
SPYGAMES is the thrilling new experience where youâll jump, climb, throw and dodge in fun immersive challenges developed with CIA experts to stretch your physical and mental agility. â Click the link below to get EARLY access as a beta tester.
SPYGAMES is coming to SPYSCAPE in Manhattan, Summer 2023
Kenya has become the first nation to ban OpenAI founder Sam Altmanâs controversial Worldcoin project, which aims to scan the retinas of the entire planet using sinister-looking 'orbs'.
Meta is reported to be launching 'personas' for their chatbots, beginning with an Abraham Lincoln bot that could be launched by next month. Other mooted personas include 'a surfer who loves to travel'.
A look at moves within the AI community to lobby for a windfall tax, which Vox writer Dylan Matthews argues would generate enough revenue to fund a global universal basic income (UBI) scheme.
Meanwhile, Nobel Prize-winning economist Joseph Stiglitz has given a fascinating interview about his fears that unregulated AI development could greatly increase financial inequality in the US and beyond.
An American Psychological Association study has found one workplace where generative AI seems to struggle; when deputizing for preachers. AI preachers generated less respect and fewer donations.
Tinder is the latest dating app to hop on the AI train, with auto-generated profile text and AI-selected photos that CEO Bernard Kim boldly claims will âeliminate awkwardnessâ for singles.
