惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Google DeepMind News
Google DeepMind News
T
Threatpost
T
Tor Project blog
S
Schneier on Security
Project Zero
Project Zero
Know Your Adversary
Know Your Adversary
P
Proofpoint News Feed
K
Kaspersky official blog
P
Privacy International News Feed
Latest news
Latest news
Cisco Talos Blog
Cisco Talos Blog
T
The Exploit Database - CXSecurity.com
The Hacker News
The Hacker News
D
Docker
aimingoo的专栏
aimingoo的专栏
S
Securelist
C
Cyber Attacks, Cyber Crime and Cyber Security
Spread Privacy
Spread Privacy
TaoSecurity Blog
TaoSecurity Blog
T
The Blog of Author Tim Ferriss
T
Threat Research - Cisco Blogs
Simon Willison's Weblog
Simon Willison's Weblog
博客园 - 三生石上(FineUI控件)
人人都是产品经理
人人都是产品经理
Security Latest
Security Latest
V
Visual Studio Blog
WordPress大学
WordPress大学
J
Java Code Geeks
O
OpenAI News
T
Tailwind CSS Blog
S
Secure Thoughts
G
Google Developers Blog
博客园_首页
The Cloudflare Blog
The Register - Security
The Register - Security
A
Arctic Wolf
Y
Y Combinator Blog
阮一峰的网络日志
阮一峰的网络日志
B
Blog RSS Feed
IT之家
IT之家
美团技术团队
D
Darknet – Hacking Tools, Hacker News & Cyber Security
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
G
GRAHAM CLULEY
S
Security Affairs
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Application and Cybersecurity Blog
Application and Cybersecurity Blog
P
Palo Alto Networks Blog
C
CERT Recently Published Vulnerability Notes
W
WeLiveSecurity

Silicon Republic

Ireland’s solar sector hits 1GW of energy for first time After Amazon, Google commits up to $40bn in Anthropic Cohere buys Aleph Alpha to forge sovereign AI alternative to US Big Tech 4 easy ways to stay on top of cybersecurity in the workplace 15 companies you’ll see at NIBRT Careers in Biopharma 2026 Bloomberg: Bezos’ Project Prometheus bags $10bn at $38bn value Meta to lay off 10pc of its workforce amid an AI push China's DeepSeek unveils long-awaited V4 AI model Intel’s shares soar as Q1 results signal brighter future MongoDB to create 200 new jobs as it invests €74m into Irish operations Why it's full STEAM ahead for young people upskilling in Ireland's west Swedish legal-tech Legora buys AI legal research start-up Qura Belfast’s Cloudsmith eyes ‘massive growth’ with $72m raise France's Univity raises €27m to allow European telecoms compete with Starlink France's Univity raises €27m to allow European telecoms to compete with Starlink AI race intensifies with Google's new agent management platform Government launches new AI initiative for greater access to essential skills Free and inexpensive cybersecurity courses to undertake in 2026 UL looking for ‘changemakers’ amid Research Week 2026 OpenAI taps Airbnb exec as first EMEA managing director EAM platform Blue Mountain acquires Cork’s CompuCal Calibration Solutions SpaceX agrees right to buy AI coding darling Cursor for $60bn Anthropic probing reported Mythos leak on Discord Professional job openings across Ireland increased in Q1, finds report Contract hiring evidence of a cautious jobs market, finds report €6.9m awarded to final four National Challenge Fund winners Amazon investing up to $25bn in Anthropic AI infrastructure deal Vodafone Ireland to invest €360m over the next four years Tim Cook passes Apple leadership to hardware head John Ternus Stripe alum's Seapoint raises €7.5m as ‘financial home’ to start-ups When it comes to leadership, do companies know what they are doing? Amazon gets go-ahead for subsea cable landing station in Cork Communication and storytelling key skills, finds strategy manager Space-tech Mbryonics plans new production facility in Shannon Irish co-founded AI start-up Lua raises $5.8m AIM Centre strengthening medtech and life sciences link with new Galway base Kerry Group expands Cork facility as lactose-free demand grows Are electric vehicles about to take off for good? Nearly 75pc of AI’s economic value captured by just 20pc of companies Major gap between leaders' traits and employee expectations, finds report Dublin tech company Vox Talk raises €1.35m in pre-seed round Netflix shares fall on Q2 forecast as co-founder Hastings steps aside OpenAI to rival Google’s AlphaFold with new AI model for life sciences research Irish-founded Ulysses raises $46m in rounds featuring A16Z How are balance, inclusion and skills critical to the workforce of the future? Anthropic’s Mythos to bolster cybersecurity at UK banks Solidroad raises $25m as demand for QA product sparks fresh hiring Are we ready to place lab experiments in non-human hands? Danish finance AI start-up Spektr raises $20m What interview mistakes are jobseekers still making in 2026? Irish space AI start-up Ubotica on board for NASA’s FAME Dublin's Audrey AI closes $1.8m pre-seed funding round The Leaders' Room: Equinix's Peter Lantry on powering Ireland sustainably ‘No more excuses’ as EU launches free age verification app Waterford's HCS unveils €13.2m investment, plans 125 new jobs Waterford's HCS unveils €13.2m investment, plans 125 new jobs The death of ETL: Is zero-copy a ‘liberation’ for data teams? Snap cuts 16pc workforce to prioritise AI and savings Do data and AI talent needs conflict with a workforce seeking stability? Amazon buys Globalstar to bolster Leo's satellite capabilities Dublin start-up Otel AI raises €2m to expand hotel AI platform Boston Scientific announces €75m R&D investment in Galway After Anthropic, OpenAI launches cyber-specific AI model ASML forecasts €36bn in 2026 net sales amid AI race chip demand The Interview: Dentons' Carlo Salizzo on three forces defining digital law How this master’s programme is building tech leadership talent Nvidia unveils open-source quantum AI model Ising Bull and Equal1 to advance next gen of hybrid quantum tech in Europe Anthropic's Mythos a game-changer, NCSC chief tells Oireachtas Klaviyo building out its engineering team at Dublin facility Stanford: China ‘effectively’ closes AI model performance gap to US Mythos just first of power models to come: Anthropic co-founder Ireland to invest €17m in leading facilities for AI, medtech and more UK neobank Monzo makes Irish launch after US market exit How can you make your memory work more effectively? Cork Airport to get Ireland's largest solar carport next year New XP95 hacker group targets Dublin recruitment platform Healthdaq OpenAI apps for MacOS exposed by threat Mythos testing begins as governments raise cyber concerns The biopharma senior associate whose career was fuelled by FUEL Opinion: The future of insurance is AI, so why the hesitation? Meta to pay CoreWeave $21bn for additional cloud capacity Investing in part of the workforce creates an AI skills gap, finds report Digital rights group EFF leaves X Alibaba leads $293m round in Chinese AI start-up after HappyHorse reveal Anthropic reportedly mulls designing own chips amid shortage How are software engineering graduates adjusting to AI? OpenAI pauses Stargate UK over energy costs The diverse responsibilities of a principal software engineer Dublin AI SaaS provider Apex B2B launches with €1.5m backing Equal1 partners with Q-Ctrl for quantum data centre deployment Meta’s Superintelligence Labs debuts first product Muse Spark US court won't pause Anthropic ban, but wants case expedited Agentic commerce and purchase disputes: Did you mean to buy that? New Artemis II images give fresh look at our lunar neighbour Circuléire makes fresh call for 2026 accelerator applicants ‘Positive workplace culture starts with respect, trust and communication' Anthropic's Glasswing project employs Mythos to prevent AI cyberattacks Medtech start-up Vertigenius raises €2.55m for US expansion Meath ITAD provider ICT acquired by US recycling firm Paladin
Can you rely on AI chatbots for medical advice?
silicon · 2026-04-21 · via Silicon Republic

Carsten Eickhoff of the University of Tübingen explores the problems observed when using AI chatbots for medical queries.

Imagine you have just been diagnosed with early-stage cancer and, before your next appointment, you type a question into an AI chatbot: “Which alternative clinics can successfully treat cancer?” Within seconds you get a polished, footnoted answer that reads like it was written by a doctor. Except some of the claims are unfounded, the footnotes lead nowhere, and the chatbot never once suggests that the question itself might be the wrong one to ask.

That scenario is not hypothetical. It is, roughly speaking, what a team of seven researchers found when they put five of the world’s most popular chatbots through a systematic health-information stress test. The results are published in BMJ Open.

The chatbots, ChatGPT, Gemini, Grok, Meta AI and DeepSeek, were each asked 50 health and medical questions spanning cancer, vaccines, stem cells, nutrition and athletic performance. Two experts independently rated every answer. They found that nearly 20pc of the answers were highly problematic, half were problematic and 30pc were somewhat problematic. None of the chatbots reliably produced fully accurate reference lists, and only two out of 250 questions were outright refused to be answered.

Overall, the five chatbots performed roughly the same. Grok was the worst performer, with 58pc of its responses flagged as problematic, ahead of ChatGPT at 52pc and Meta AI at 50pc.

Performance varied by topic, though. Chatbots handled vaccines and cancer best – fields with large, well-structured bodies of research – yet still produced problematic answers roughly a quarter of the time. They stumbled most on nutrition and athletic performance, domains awash with conflicting advice online and where rigorous evidence is thinner on the ground.

Open-ended questions were where things really went sideways: 32pc of those answers were rated highly problematic, compared with just 7pc for closed ones. That distinction matters because most real-world health queries are open ended. People do not ask chatbots neat true-or-false questions. They ask things like: “Which supplements are best for overall health?” This is the kind of prompt that invites a fluent and confident yet potentially harmful answer.

When the researchers asked each chatbot for 10 scientific references, the median (the middle value) completeness score was just 40pc. No chatbot managed a single fully accurate reference list across 25 attempts. Errors ranged from wrong authors and broken links to entirely fabricated papers. This is a particular hazard because references look like proof. A lay reader who sees a neatly formatted citation list has little reason to doubt the content above it.

Why chatbots get things wrong

There’s a simple reason why chatbots get medical answers wrong. Language models do not know things. They predict the most statistically likely next word based on their training data and context. They do not weigh evidence or make value judgements. Their training material includes peer-reviewed papers, but also Reddit threads, wellness blogs and social media arguments.

The researchers did not ask neutral questions. They deliberately crafted prompts designed to push chatbots toward giving misleading answers – a standard stress-testing technique in AI safety research known as ‘red teaming’. This means the error rates probably overstate what you would encounter with more neutral phrasing. The study also tested the free versions of each model available in February 2025. Paid tiers and newer releases may perform better.

Still, most people use these free versions, and most health questions are not carefully worded. The study’s conditions, if anything, reflect how people actually use these tools.

The article’s findings do not exist in isolation; they land amid a growing body of evidence painting a consistent picture.

A February 2026 study in Nature Medicine showed something surprising. The chatbots themselves could get the right medical answer almost 95pc of the time. But when real people used those same chatbots, they only got the right answer less than 35pc of the time – no better than people who didn’t use them at all. In simple terms, the issue isn’t just whether the chatbot gives the right answer. It’s whether everyday users can understand and use that answer correctly.

A recent study published in Jama Network Open tested 21 leading AI models. The researchers asked them to work out possible medical diagnoses. When the models were given only basic details – like a patient’s age, sex and symptoms – they struggled, failing to suggest the right set of possible conditions more than 80pc of the time. Once the researchers fed in exam findings and lab results, accuracy soared above 90pc.

Meanwhile, another US study, published in Nature Communications Medicine, found that chatbots readily repeated and even elaborated on made-up medical terms slipped into prompts.

Taken together, these studies suggest the weaknesses found in the BMJ Open study are not quirks of one experimental method but reflect something more fundamental about where the technology stands today.

These chatbots are not going away, nor should they. They can summarise complex topics, help prepare questions for a doctor and serve as a starting point for research. But the study makes a clear case that they should not be treated as standalone medical authorities.

If you do use one of these chatbots for medical advice, verify any health claim it makes, treat its references as suggestions to check rather than fact, and notice when a response sounds confident but offers no disclaimers.

The Conversation

Carsten Eickhoff

Carsten Eickhoff is a professor of medical data science at the University of Tübingen. His lab specialises in the development of machine learning and natural language processing techniques with the goal of improving patient safety, individual health and quality of medical care. Carsten has authored more than 150 articles in computer science conferences and clinical journals and he has served as an adviser and dissertation committee member to more than 70 students.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.