Hear! Hear! The Dangerous Illusion of the “Two-Week AI Expert”: A Call to Action Against Superficial Medical Education
––– 凄い –––
This is so true! Doctors need to know about the artificial intelligence tools they use. Right now, they are at the mercy of engineers who think they know how to implement AI in the clinic. So few actual doctors are involved in the design or decision-making.When a short-course provider cuts out the fundamentals — claiming that logic, data architecture, or algorithmic limits are “too complex” or “too theoretical” — they are not saving clinicians time. They are stripping away their critical armor. When you do not understand why a probabilistic model produces an answer, you cannot evaluate when that answer is a dangerous hallucination.
What is the result of this shortcut culture? A passive acceptance of systemic failure! We hear healthcare workers repeating slogans: “Well, that’s just AI; it hallucinates sometimes, it’s a black box, and it has bias.”
We have normalized the error before we even bothered to teach the mechanism! Why have we accepted that clinicians should blindly trust — or helplessly dismiss — a tool simply because we were too lazy to teach them how it actually works?
Tesla to pull the plug on Cybertruck after years of poor sales. From early corrosion problems to its weird ugly design, I'm not surprised that sales were poor.
––– 良くない–––
Open since 2008, the conjecture of Scott Aaronson and Andris Ambainis concerns when quantum computers can be exponentially faster than ordinary ones. Now, a research system developed internally at Physical Superintelligence (PSI), running a non-public Anthropic model, found a proof for it, settling one of the most fundamental questions in quantum computing.
The new Godzilla movie is incredibly loud and may impair your hearing. I'll wait for the streaming release.
––– 良くない–––
12,600-year-old painting that's 8 miles long discovered in the Amazon Rainforest. 12,600 years??? What is that pigment made of!
––– 凄い –––
Anthropic's unreleased Model 4, trained in September–October 2026, scored 178.6 on the AI capability index, a jump nearly 90% as large as the entire improvement from Opus 4.6 to Mythos 5. Wow.
And in the open-weights world:
Thinking Machines has released Inkling 2 Small, scoring 31 on the Artificial Analysis Intelligence Index - their plan to release its weights would make it the highest-scoring open weights model developed outside China based on current rankings.
Is Google in trouble? Their stock hasn't been doing well that's for sure. They haven't been attracting the brains as much as OpenAI and Anthropic. They've gotten wokey and DEI. Are they becoming like Microsoft in 2010? And Warby Parker seemingly spilled the beans on Google's Android XR smart glasses launch. Ugh, just what we don't need – another creep glasses like Meta's.
And Microsoft is trying to pay people to leave. That doesn't sounds like a growing company.
––– 良くない–––
Math research is surging. Biology isn’t. The benefits of AI are not evenly distributed. That's because biology is not symbolic manipulation like math is. AI in biology is more stuff like data analysis, vision transformers and autoencoders and Restrictive Boltzmann Machines, not so much LLMs. But give it time.
––– 凄い –––
Lockdown Mode Bypassed in New iPhone Exploit. Security is never a certainty with Apple. It's a consumer product, after all.
––– 良くない–––
Breath Harbor offers free breathing, grounding and meditation tools. I've never favored meditation apps where you have to purchase with your Apple ID or Google ID, or login to a website. These apps are free and require no log in or payment. Truly a public service. All you really need is just something to help you get your attention back to breathing when it wanders. Doing so, reinforces the anterior cingulate gyrus and helps to tamp down activity in the amygdala and locus ceruleus. Regular practice is the key.
––– 凄い –––
Apple wants you to switch to Safari. I only use Safari when I really need to access something my privacy-enable LibreWolf browser can't access the webpage, usually because it requires that I accept fingerprint revealing information. Or if I really want to use Apple Pay. I wouldn't "switch" to Safari. Just use it occasionally when necessary.
––– 凄い –––
Verify, Then Vote. I've always been skeptical of the mixture of experts models which accept as the correct answer the majority vote from either different models or different inference results using the same model. How do you know that an answer is correct. A majority vote over wrong answers is still wrong. One should verify. But that takes extra time and effort, and is more like a search engine. This blog post is just a pilot study, but really, it's what models should be doing.
––– 凄い –––
OpenAI Just Found Nearly 1,000 Ways To Gain Root Access In Linux. This is why we should use AI in coding. It's far superior to the human-alone coding we've been doing. But humans still need to be involved.
––– 凄い –––
Another report that Claude Agents will leak your email address to third parties. Using autonomous agents is still risky. Too much for me.
––– 良くない–––
Novartis confirms 9 reported cases of tumors in patients treated with Zolgensma. Out of over 6000 treated. This is an FDA approved gene therapy treatment for spinal muscle atrophy.
––– 良くない–––
Here's a website that tells you What has AI actually done? I guess it has done some good things.
––– 凄い –––
What causes mental fatigue in the brain? A build up of glutamate in the lateral prefrontal cortex. Maybe now we can start to work on ways to reduce this. Glutamate is toxic if it builds up.
––– 凄い –––
Multnomah County approves ambulance service plan that sets longer response goal. Really too bad, Fairview, Troutdale and Wood Village. Next time, don't re-elect Democrats.
––– 良くない–––
