Microsoft claims its AI medical tool can outperform human doctors in diagnosing complex cases
Called the Microsoft AI Diagnostic Orchestrator (MAI-DxO), the system mimics a panel of five virtual doctors, each with its own specialty, working together to debate and diagnose.
In trials, it solved nearly 86 per cent of difficult medical cases published in The New England Journal of Medicine, which is known for complexity.
That is over four times more accurate than experienced human doctors under the same conditions, according to the tech's co-founder Dr Mustafa Suleyman.
'We're talking about models that are not just better, but dramatically better — faster, cheaper, and more accurate than human doctors,' Suleyman said.
Experts urge caution — it's early days.
The system made the diagnoses using patient data already provided, and clinical use is still a way off, with peer-review and safety testing ahead.
University of Sydney biomedical informatics professor Adam Dunn said it is not a real representation of what doctors do.
'These tools can be useful as an assistant, but they're never going to replace doctors,' Dunn told 7NEWS.
'The doctor has to quickly look at you and understand what kinds of questions to ask you to gather the right kinds of information.
'The second important thing to think about for ethics is related to who the AI is intended to support.
'In a lot of cases if the data are from one hospital in Boston, it's not a representative population.
'The decisions that an AI can make may be unfair. They may not be protecting our most marginalised and vulnerable populations.'
Artificial intelligence is currently being used across healthcare in many areas, with one standing out in terms of popularity.
Known as AI Scribes, the system listens to the conversation between a patient and a doctor or healthcare provider, and then quickly summarises the chat for medical records.
Australian Medical Association chair of public health Dr Michael Bonning said it allows the clinician to spend more time with the patient and less on paperwork.
'We know, and many practitioners will say this, that because of the time pressures on their day, they will often write a short series of notes and then rely on their memory of the consultation as well to provide further insight down the track,' Bonning told 7NEWS.
'In a six or eight-hour day, you might spend an hour and a half of that typing. And if in that circumstance you can let the AI do that, that's a huge time saving for a clinician.'
Doctors still review the transcription.
'Cannot replace humans'
Members of the public told 7NEWS they welcome the use of artificial intelligence as a tool to assist doctors, but would not trust the computer in place of a human doctor.
'I think it can be used as a support b, butot completely replacing humans,' one person said. 'I think we still have to have human reviewing in there.'
Another said a 'human doctor is still better, but I guess everything is artificial intelligence'.
'It's scary, but at the same time you don't know what the future holds with AI. It's both good and bad.'
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
7NEWS
7 hours ago
- 7NEWS
Melbourne family claims to find lizard inside can of cannellini beans
A Melbourne family claims to have found a dead lizard inside a can of beans purchased from their local supermarket. Harrison Razzi and his family were preparing one of their favourite meals when they made the horrifying discovery on the weekend. WATCH THE VIDEO ABOVE: Family finds lizard in canned beans, sparks investigation. Razzi said the family was making minestrone to 'have a good food, a good dinner'. But when he opened a can of Carmelina's cannellini beans, Razzi said he found a small lizard inside. 'I was looking in the can of beans and found a lizard, a baby lizard in there,' he said. 'My mum was crying, my sister was screaming.' Razzi said the sight was 'horrifying' and it looked like the lizard had been fermented in the can of beans. The importer of the beans, Leo's Imports, told 7NEWS it 'takes all consumer complaints seriously and is committed to upholding the highest standards of product quality and safety'. However it added: 'We are unable to substantiate the claim which raises questions regarding its credibility and intent.' The company says it has committed to a full and proper investigation. According to the product's packaging, the beans come from Italy. It is not the first time an animal has made its way into a tinned product. Last year a South Australian woman claimed she found the remains of rat in a Coles brand can of red kidney beans. Melbourne University food scientist Senaka Ranadheera said it was common for amphibians to be found in fresh produce. 'They have previously found even birds or small mammals like mice,' he said. 'We need to further improve these food safety management systems.' Razzi said his family has been buying the same brand of Italian beans for years without issue. But he said they are not in a rush to buy another can and attempt to make minestrone again. 'Maybe in the near future but not anytime soon,' he said.
West Australian
9 hours ago
- West Australian
Microsoft hit with SharePoint attack affecting global businesses and governments
Microsoft has warned of 'active attacks' targeting its SharePoint collaboration software, with security researchers noting that organisations worldwide stand to be affected by the breach. The Cybersecurity and Infrastructure Security Agency said Sunday in a release that the vulnerability provides unauthenticated access to systems and full access to SharePoint content, enabling bad actors to execute code over the network. CISA said that while the scope and impact of the attack continue to be assessed, the agency warned that it 'poses a risk to organisations.' Microsoft late Sunday issued fixes for customers to apply to two versions of the SharePoint software. Another 2016 version remains vulnerable and the company said it is working to develop a patch. Researchers at Palo Alto Networks said the hack likely reached thousands of organisations globally. 'The exploits are real, in-the-wild and pose a serious threat,' they added. A Microsoft spokesperson declined to comment on the incident beyond what was shared in a company blog post. In an alert Saturday, Microsoft said the attack applies only to on-premises SharePoint servers, not those in the cloud like Microsoft 365. SharePoint software is commonly used by global businesses and organisations to store and collaborate on documents. The vulnerability is especially concerning because it allows hackers to impersonate users or services even after the SharePoint server is patched, according to researchers at European cybersecurity firm Eye Security, which said it first identified the flaw. SharePoint servers often connect to other Microsoft services such as Outlook and Teams, meaning such a breach can 'quickly' lead to data theft and password harvesting, Eye Security researchers said. 'Once inside, they're exfiltrating sensitive data, deploying persistent backdoors, and stealing cryptographic keys,' Michael Sikorski, CTO and head of threat intelligence for Palo Alto's Unit 42, said in a statement. 'The attackers have leveraged this vulnerability to get into systems and are already establishing their foothold.' Separately, Alaska Airlines briefly halted its ground operations for about three hours on Sunday due to an IT outage. It lifted the ground stop at roughly 2 a.m. EST, the carrier said in a statement. It was unclear whether the outage was related to the SharePoint attack.
West Australian
13 hours ago
- West Australian
Microsoft server hack hit 100 organisations: analysts
A sweeping cyber espionage operation targeting Microsoft server software has compromised about 100 different organisations, two of the groups that helped uncover the campaign say. Microsoft on Saturday issued an alert about "active attacks" on self-hosted SharePoint servers, which are widely used by organisations to share documents and collaborate within organisations. SharePoint instances run off of Microsoft servers were unaffected. Dubbed a "zero-day" because it leverages a previously undisclosed digital weakness, the hacks allow spies to penetrate vulnerable servers and potentially drop a backdoor to secure continuous access to victim organisations. Vaisha Bernard, the chief hacker at Eye Security, a Netherlands-based cybersecurity firm, which discovered the hacking campaign targeting one of its clients on Friday, said that an internet scan carried out with the Shadowserver Foundation had uncovered nearly 100 victims altogether - and that was before the technique behind the hack was widely known. "It's unambiguous," Bernard said. "Who knows what other adversaries have done since to place other backdoors." He declined to identify the affected organisations, saying that the relevant national authorities had been notified. The Shadowserver Foundation confirmed the 100 figure and said that most of those affected were in the United States and Germany and that the victims included government organisations. Another researcher said that, so far, the spying appeared to be the work of a single hacker or set of hackers. "It's possible that this will quickly change," said Rafe Pilling, director of Threat Intelligence at Sophos, a British cybersecurity firm. Microsoft said it had "provided security updates and encourages customers to install them," a company spokesperson said in an emailed statement. It was not clear who was behind the ongoing hack. The FBI said on Sunday it was aware of the attacks and was working closely with its federal and private-sector partners but offered no other details. The UK National Cyber Security Centre said in a statement that it was aware of "a limited number" of targets in the United Kingdom. A researcher tracking the campaign said that the campaign appeared initially aimed at a narrow set of government-related organisations. The pool of potential targets remains vast. According to data from Shodan, a search engine that helps to identify internet-linked equipment, more than 8000 servers online could theoretically have already been compromised by hackers. Those servers include major industrial firms, banks, auditors, healthcare companies and several US state-level and international government entities. "The SharePoint incident appears to have created a broad level of compromise across a range of servers globally," Daniel Card of UK cybersecurity consultancy PwnDefend said. "Taking an assumed breach approach is wise, and it's also important to understand that just applying the patch isn't all that is required here."
