Brian White

Killing human beings is never anything to make light of – particularly when it comes from someone who has the most powerful military in the history of the world at his disposal. This is beyond obscene.

Trump’s night of 160 posts on Truth Social fuels debate about US president’s stamina The 79-year-old president’s social media blitz included conspiracies, attacks on political foes and self-praise

Maybe he used the AutoPost

Democrats should roll these videos on the House and Senate floors.

Saying the admiral was “thrown under the bus” makes him sound like a victim. He knowingly followed an illegal order. The SecDef gave it. Both are culpable. It highlights a leadership culture where illegal orders are willingly given and carried out, then laundered through public excuses.

It is increasingly clear that an enfeebled Trump isn’t leading. He claims no knowledge of who he’s pardoning, no knowledge of military strikes, broader administration policy, etc. He’s effectively a human autopen for oligarchs and fascists who fill the vacuum and manipulate him to sign things.

The percentages are small in terms of imports by China and India, but the numbers are likely large enough to be a measurable boost to Russia. And it allows Trump to maintain a veneer of sanctions, while boosting Russian oil income.

There’s no doubt other sinister angles. But, you know, Russia.

So if a Venezuela operation just happens to funnel more money into Russia’s sanctions-dodging exporters, the question writes itself: who actually benefits most from this move?

With Trump, everything always points to Russia.

A Venezuelan shortfall tightens supply, raises prices, and turns Russia’s shadow fleet into an even larger revenue engine.

A “drug mission” that results in a wartime windfall for Moscow is…quite a coincidence.

And Russia needs higher margins. Putin’s war machine is burning through cash as fast as it can refill it. Every extra dollar per barrel goes straight into sustaining the invasion of Ukraine.

Venezuela ships roughly 800k barrels/day. Knock that offline and you instantly create a global supply gap for China and India — one that Russia’s shadow fleet is perfectly positioned to fill.

They already move ~70% of Russia’s seaborne crude. A Venezuela shock would just boost their margins.

Other, apparently more morally developed countries, deal with this stuff appropriately.

Very sorry you are going through that. Wishing you the best.

So, if one is asking about medical issue as in your example, perhaps ask it to cite and link to reliable sources like the Mayo or Cleveland Clinics, or JAMA or The Lancet. Then you can click through to those sources and validate the response.

That is definitely one of the issues with current models when prompted for that type of technical response. When it comes to technical questions and topics, I find it helpful to enable a “research mode” if the model you are using has that feature, or simply including a prompt to cite sources.

I’ve used several standard LLMs in my work, and have seen solid efficiency and productivity gains when used thoughtfully and in the proper context. Like any tool, the user has the primary and ultimate responsibility for how the output is evaluated and used.

And seriously — thanks for the thoughtful, civil back-and-forth. Conversations about this stuff usually melt down fast, and it’s great to actually have non-inflammatory dialogue about it.

And this is also where it’s different from a simple algorithm. An algorithm just follows fixed rules. These models learn the rules from huge datasets, adapt to new examples, and improve with retraining. Still not human intelligence, but definitely not a static script either.

Yeah, that’s exactly it — the “intelligence” in AI is artificial. It doesn’t have intent or curiosity, but it does learn patterns far beyond what humans can see.

It won’t decide to detect CO₂ on its own. A human sets the goal, and the model learns how to hit it. That’s the partnership.

AI that automatically detects methane plumes from space could be a University of Oxford researchers, in partnership with Trillium Technologies’ NIO.space, have developed a tool to automatically detect methane plumes on Earth from orbit using machine learning with

In this case researchers are using actual deep-learning models. This methane-mapping work isn’t a just simple algorithm. It uses AI models trained on hyperspectral data. The system auto-detects methane plumes from orbit. Here an article from the University of Oxford.

There’s a link in this article to a study in Lancet Oncology, in which 80k mammograms were evaluated – half by radiologists, half by AI. The AI group had 20% more cancers detected. Another study with over 1 million mammograms that used AI along with a radiologist outperformed the radiologist alone.

I agree there are real debates about the social and economic impacts. But it’s not true that “no one says AI is bad” — plenty of prominent critics say exactly that. And the breakthroughs I cited didn’t “just happen with a laptop and a chair.” AI meaningfully expanded what humans could do.

I’m dismayed by many smart people who paint AI with a broad brush. It often is reduced to some selective, poor examples of output produced by models like ChatGPT. AI is a category of tools, and in many instances, they are being used effectively for important, even groundbreaking, scientific work.

I have huge respect for Applebaum’s scholarship and writing, but am disappointed when serious academics use such a broad brush, and are dismissive of an entire category of tools, seemingly because some have produced poor quality output. There are many breakthroughs that improve life for all of us.

AI is much more than someone’s poor use of ChatGPT. Advanced AI models have unlocked protein-folding predictions that would have taken scientists decades, sped mRNA vaccine design, found new ALS and cancer drug pathways, and has radically improved environmental modeling.

Just saying, “AI is bad” has the same energy as claiming “books are dangerous” because someone misread a cookbook. I think the discourse deserves more rigor and balance.

NOTE: This isn’t specifically aimed at Matt. Just reflecting on the example he posted, and some comments to it . 9/9

And AI-driven satellite analysis is now mapping methane leaks in real time. It’s already flagged major oil-and-gas emitters, enabling faster regulatory action. 8/9

Environmental science, too: AI climate models are improving extreme-weather forecasting, catching localized flood and heatwave patterns traditional models miss. 7/9

MIT used AI to identify new antibiotic candidates. These are compounds that kill drug-resistant pathogens humans couldn’t find. 6/9

AI models in radiology now detect breast cancer earlier than human-only reads. This reduces false negatives, speeds intervention, and saves lives. 5/9

DeepMind’s AlphaMissense mapped thousands of disease-causing genetic variants. Before that, researchers spent years guessing which mutations actually mattered for cancer, neuro disorders, and rare diseases. 4/9