A more accurate alternative to anxious Googling of symptoms was announced at Google’s IO event.
There is a certain fear that comes with attempting to diagnose your own symptoms via Google, ruling out dangerous diseases and unheard-of plagues in the search for the common cold you’re facing. But Google is looking to change this, replacing your GP with a friendly artificial intelligence (AI) platform.
As part of Google’s IO event, a chance to unveil the company’s biggest projects of the year, a program known as ‘PaLM 2’ was revealed. Like OpenAI’s GPT-4 (the technology behind ChatGPT), PaLM 2 is a powerful language model, able to deal with code, reasoning, multiple languages and huge amounts of text.
While language model technologies are being crammed into everything from cybersecurity to web search engines and the ChatGPT competitor Google Bard, one of its most advanced uses could be in the medical field. Med-PaLM 2 is the company’s second version of an AI platform designed specifically to provide answers to medical questions.
![""]("https://images.immediate.co.uk/production/volatile/sites/4/2023/05/med-palm-dc4977b.png?quality=90&resize=620%2C300" "\"\"")
When tested on the United States Medical Licensing Examination (USMLE), the model scored 85.4 per cent, in line with the knowledge of a medical expert.
To turn this model into an expert, Google trained it on… well, other experts. A panel of clinicians from the UK, US and India were given a long set of medical questions and scenarios. The model was then trained to answer questions in a similar way to these experts.
The clinicians then cross-referenced the model’s answers on a set of values including a low likelihood of medical harm, alignment with scientific consensus, precision and a lack of bias.
Of course, while an AI model can understand the logic of science and medicine, it isn’t exactly great on the ethical or moral side of things. Further training was done to help the model align with human ethical values in an attempt to improve its bedside manner.
This isn’t the first time an AI platform has played doctor. This is the second iteration of Google’s technology and OpenAI has also passed the USMLE test. This latest version from Google is, however, the first model to pass at the level of an expert.
While it is certainly impressive, doctors don’t need to worry about losing their jobs just yet – this technology is still in its earlier stages. Google will first open up the model through Google Cloud, looking to get users’ feedback on its ability.
In the future, further capabilities could be added. Right now, the model is limited to an understanding of text, but future versions could understand your medical record, CT scans, or genomics data.
Read more:
- ChatGPT is evolving. But is it smart enough to have a place in our future?
- How ChatGPT – and Ted Lasso – could rid the internet of hate speech
- AI: 5 of the best must-read artificial intelligence books