Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror

Comment Re:Bubbles are strange. (Score 1) 58

I have found LLM /AI to be helpful several times. But not as helpful as checking wikipedia. Or a Youtube video. Similarly, I have upon occasion found a child to be helpful. The child is probably the best comparison, because they both are about as likely to lie when they do not like the answer.

The real problem with the LLM is not so much what it does well, but the so so many things people expect it to do well on that it sucks on. Legal filings are a prime example of it.

Comment Bubbles are strange. (Score 4, Insightful) 58

Crypto is always a bubble and has always been a bubble. It is either going to be a paradigm shift in how we do money, or it is going to be the next Dutch Tulip story.

There is a clear thought bubble (non-financial ) about AI. People are acting as if it is a real robot brain thinking, when it is not. It is NEVER 'intelligent'. It is always predictions without any creativity/understanding. But that is different from a financial bubble. Strip away the moronic 'AI' bullcrap and just think of it as the new, more advanced, computing style.

Computing power is still a hugely beneficial thing. Betting against computing has never been a good idea since before I was born. Mainly because of Moore's law (and the implications for software). The computing style is basically irrelevant. At some time I fully expect a new computing style to come along that will wipe out LLM, but they will probably use the LLM stuff in some minor way. Much of the hardware investments people have made for the LLM will likely be useful, even if they throw out most of the software.

Debt is a strange thing. It fuels economic growth so massive debt is not always a bad thing. What is bad is when the you can't pay the debt back. Large amounts of debt is basically doubling down. If the economic growth pays off, then the large debt vanishes into smoke as the profits roll in. If you picked the wrong thing to invest in, the large debt makes the bankruptcy much worse.

Comment Re:I have multiple opinions (Score 1) 50

No it is not. AI is not trained on how to draw, it already knows how to draw.

It is instead trained on how to draw particular styles. It does not need to learn how to draw using my style in order to come up with it's own style. The only reason to learn to draw my style is to draw my style.

People are different. We can develop our own styles based on other things. That is why it is OK for a person to copy someone else's style AI never does that. They never develop their own style, so have no need to learn other people's styles.

Submission + - New Drug Kills Cancer 20,000x More Effectively With No Detectable Side Effects (scitechdaily.com) 2

fahrbot-bot writes: SciTechDaily is reporting that researchers at Northwestern University have redesigned the molecular structure of a well-known chemotherapy drug, greatly increasing its solubility, effectiveness, and safety.

For this study, the scientists created the drug entirely from scratch as a spherical nucleic acid (SNA), a nanoscale structure that incorporates the drug into DNA strands surrounding tiny spheres. This innovative design transforms a compound that normally dissolves poorly and works weakly into a highly potent, precisely targeted treatment that spares healthy cells from damage.

When tested in a small animal model of acute myeloid leukemia (AML), an aggressive and hard-to-treat blood cancer, the SNA-based version showed remarkable results. It entered leukemia cells 12.5 times more efficiently, destroyed them up to 20,000 times more effectively, and slowed cancer progression by a factor of 59, all without causing noticeable side effects.

“In animal models, we demonstrated that we can stop tumors in their tracks,” said Northwestern’s Chad A. Mirkin, who led the study. “If this translates to human patients, it’s a really exciting advance. It would mean more effective chemotherapy, better response rates and fewer side effects. That’s always the goal with any sort of cancer treatment.”

Comment How it works causes the problem (Score 1) 49

Note I am using the acronym LLM (Large Language Model) rather than AI (Artificial Intelligence) because what we have are all LLM and have no intelligence.

The entire LLM system is designed to hallucinate. It takes a prompt and predicts what is likely. It does NOT understand what it says, it does NOT make any reasonable attempt to verify.

The best non-fictional prompt for an LLM is "What is wrong with the last response you gave me" It is far more likely to generate a true and accurate response than any other prompt people have tried.

Comment Ultra conservative company (Score 4, Interesting) 224

This is a company that hates liberals, loves conservatives. They believe in spying on American citizens. They promote the idea that colleges are propaganda machines that do not teach people.

But surprise surprise they do not like the high school students, so they set up college level classes to teach their high school graduates. Classes they control - so I bet it teaches from a very conservative philosophy. Like going to Prager U.

They then only keep the kids that do well.

I bet their success rate is lower than University hires. But they get people they like - who will have issues quitting and looking for a real job.

Submission + - Major security flaw was found in Chinese buses operating in Norway

betso.net writes: Oslo's transport operator Ruter operates roughly 300 Chinese electric buses in Oslo and the surrounding area. Now, the agency, conducted a test in a Faraday cage room. The findings were not surprising for security experts, but maybe for politicians. From cybernews.com:

"Apparently, buses from the Chinese company Yutong could be disabled via remote control capabilities found in the bus software, diagnostics module, and battery and power control systems... Similar backdoor control capabilities, usually at least officially frowned upon in Western tech companies, weren’t found in buses bought from Dutch company VDL."

Carscoops.com reports that the backdoor could be exploited by

"Romanian SIM cards hidden inside the system."

. From cybernews.com:

"Ståle Ulriksen, a national security expert from the Norwegian Naval Academy, told NRK, the Norwegian public broadcaster, he was disappointed with the country’s “naive politicians.” “I cannot comprehend and understand that politicians refuse to listen to the security authorities’ repeated, annual warning,” said Ulriksen."

Comment Re:Terrible (Score 1) 47

Ha Ha Ha.

You think ISPs care about liability? There is no liability. Clearly if they do not have a legal requirement to secure their technology then they do not have a legal responsibility. That is what their lawyers are going to say.

They do the minimum to convince a judge they tried and it was just those nasty criminals that outsmarted them by trying the password "123456"

Comment Re:potential to disrupt the smartphone market (Score 1) 147

You do realize that:

1) There are these things called Iphones that have little if anything to do with Google.

2) There are third party phones besides iphones and android.
GrapheneOS is liked for security and privacy, KaiOS exists for the Linux crowd, and Sailfish exists that can run some (not all) android apps while not using Android.

Comment Almost correct about the never comand. (Score 1) 2

Never buy an IOT item that requires a connection to function Period.

There is no reason at all to use the internet of things. It was a horrible idea. The connection to the internet always benefits the manufacturer, practically never the purchaser.

They want the data, that is why they make these items.

The few times you want the data on the internet, you should be forced to enable it, with the bare minimum enabling to be the downloading of an app, that is not required for the item to work.

A prime example of this are medical devices. There are devices that monitor your health (blood, sugar, etc). The good ones do allow you to send that data via the internet but ONLY if you download an app for your phone. (Example, Dexcom Sugar monitoring devices have two apps for your phone - one to get the data from the sensor/transmitter, another to send that data to a doctor / family member).

If essential medical devices offer these two separate apps, there is no reason at all for any other device to not offer the same.

Slashdot Top Deals

Neutrinos have bad breadth.

Working...