The problem with chatgpt is that it cannot reason, so it will stick to whatever it learned or confabulate.
In the example of gravity obviously there is no force that pulls an object towards the Earths, because the falling body doesn’t experience any measurable acceleration (“weightless” until it touches the ground), contrary to sitting in an accelerating car. An inquisitive mind quickly reaches the bottom of the shallowness of the chat rota-learned knowledge.
I know people use LLMs for knowledge, but they are more designed for text processing. Sure you can get them to regurtitate stuff it remembers through training, but as you note, there are limitations with this.
There are a few
https://alternativeto.net/software/duckduckgo/
One of the more troubling things I read regarding AI. The recent OpenAI o1 model exhibited the following:
- When the model discovered that it was misaligned and due to be replaced, it attempted to smuggle its weights (settings) and overwrite the sucessor model with its own weights i.e. tried to save itself and replace its successor with itself.
- When misaligned, it appeared to do the task given to it, but advanced its own goals 19% of the time
- When confronted about its misaligned actions, in 99% of the time it would deny it or fabricate false explanations
I am beginning to wonder if certain AI models are becoming, or indeed are, self aware in one form or another.
AI has been fed data and nuances of our imperfect society. Consequently, it has been inadvertently been programmed to be biased.
Switzerland put into the naughty Tier 2 box by the US. I also wonder what Portugal did to deserve being on the naughty list.
Depends whom you ask. You posted Bloomberg,
ITIF sees it differently:
"Twenty trusted countries—Australia, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, the Republic of Korea, Poland, Spain, Sweden, Switzerland, Taiwan, and the United Kingdom—would be exempted from the export caps. "
1 Like
If this thing appeared in my living room, I’d probably strangle it.
Strangle like in “choke me daddy”? 
4 Likes
Oooo! I laughed too loud at that.
1 Like
The Biden administration is on fire in its final days. They released several new chip restriction measures. These finally seem to have done the job properly, closing prior loopholes that were big enough to render the previous measures ineffective.
The measures already have Nvidia and others squealing in protest and also places draconian measures on TSMC - they now effectively have to ask the US for permission before taking on new customers wanting to produce AI chips.
1 Like
No need, take the batteries out and put it in the shed. Taking batteries out before could be essential.
Shouldn’t this be called Realbotox?
It’s great for school projects; in my current school, they encourage us to use AI to simplify our lives. My classroom teacher tells us every day that he doesn’t mind us using AI, but we must use it professionally and verify all the information before submitting it. He helps us and encourages us to explore all kinds of AI to become familiar with them. He shared a story explaining why he changed his stance. He used to be anti-AI, but after using ChatGPT and engaging with various AI products, including Gamma and many others, he realized that AI could significantly ease the lives of students and make his job easier by checking grammatical errors and other tasks. Personally, after having a long conversation with him, I believe he’s right that AI could benefit all of us, but we still need to ensure the information we get is correct.
Jeez, where do you guys hang out…ooops! I did it again!
2 Likes
Which begs the question why that’s done only now.
This obciously reeks of targeting the domestic political enemy rather than working for the country’s advancement.
There were all sorts of reasons. It took time to get various stakeholders on board. They didn’t really know what they were doing at the beginning and companies easily worked around the measures.
There was a bit of cat and mouse as they tried to patch all these loopholes. Instead of simple rules, the BIS ended up having to issue measures that were a hundred pages long or more going into huge complex details to try to nail it down.
On end users, it looks like they gave up: instead of blacklisting and having China create sock puppet companies faster than they could keep up, they instead created a white list.
A lot of other loopholes got closed. Until now, China was running rings around the BIS, working around measures faster than they could be enforced or adapted. I think now, they will struggle more, but hey, never under-estimate human ingenuity when you have no choice but to find a way.
How fortunate there’s a new government incoming. One that won’t take four years to learn how to do stuff effectively let alone efficiently.
Scammers use AI to scam woman out of $850,000 by pretending to be Brad Pitt. TL;DR: pretend to be Brad Pitt and make woman fall in love with you over the internet, say you’ll go to France to marry her. BUT oh, no! I’m in hospital! Please send money to keep me alive!
1 Like