Is this like how you would praise the 99% of people who walk by your house without stealing something ... because 99 >> 1, so we can praise the 99 and not harp on the 1?

Is this like the difference when one's boss "asks" versus "insists" or "orders"? When a superior "asks," it's not the same as when a peer "asks."

What does "competitive" mean? If we're talking stock price appreciation, then Nvidia and AMD are leaving Intel in the dust. If we're talking benchmarks or reviews or Slashdot fandom, Intel is once again behind. If we're talking market share, then AMD has never been competitive with Intel, in any CPU segment. It's not even close. In fact, if we're talking market share and not benchmarks for GPUs, AMD hasn't been close to competitive with Nvidia for a decade. In terms of GPU market share, AMD is much closer to Intel than Nvidia.

The problem with leaking information has nothing to do with traditional security issues. The problem is that any query can become data that modifies or trains the model.

All sane companies already have policies about what can be included in an external gen AI query. For example, generic code questions are fine, but copying any portion of any proprietary code is forbidden.

These points could all be applied to human writers. Human writers can hallucinate every bit as well as AI engines by misremembering things and by referencing incorrect sources. Furthermore, human writers can outright lie and have bias-driver agendas. The editor does not necessarily have an easy job, but it's not obvious that the editing job is harder with ChatGPT. In fact, this would be an interesting research experiment.

Subjective age was based on asking study participants this question: “On some days you may feel older or younger than your calendar age. What age do you feel right now?”

This seems like a totally worthless question to me. What's the precision on the expected answer? Maybe I feel like 35.3 years now, but yesterday I felt like 35.8 years? My precision would be on the order of decades. And my scale would change from hour to hour depending on my mental and emotional state. Asking for for subjective age seems similar to asking for subjective happiness. Perhaps there is some significance for how one specific person feels from one moment to the next, but I'm skeptical that a quantitative guess has a non-zero significance across multiple individuals.

Fair points, but the same could be said for textbooks. Neither textbooks nor ChatGPT (and let's throw stackexchange in for kicks) will help clueless students. All these things are resources. Expecting more (such as solving entire problems rather than pointing the way to limited solutions for specific parts of the problem) is a problem with expectations and not necessarily the resource.

The difference between #1 and #2 is 2 points, with a confidence interval of at least +/-4. That is, the two are statistically similar.

Of course, that's even assuming that a competition based on crowd-sourced human preference votes means something. If anything, this contest measures popularity, but it probably means much less than that due to methodology holes.

"We followed all BBC editorial compliance processes and the final text was verified and signed-off by a member of the marketing team before it was sent."

This is anti-AI hysteria. Having a human proofread and approve AI-generated content is exactly the proper way to use generative AI. It is no different than having a junior writer create something that is proofread and approved by a human editor.

My field of specialty is computer fault tolerance, and I've never heard of "empathy" used for fault tolerance. In fact, at least based on the linked articles, it's not even clear what "empathy" means. However, what the article describes about algorithms for probing systems and determining when to repair and when to give up sounds quite conventional. The only innovation that I can see is the invention of the term "empathy."

Of course, there are near-term effects. AAPL dove 3.8% in the four hours after the lawsuit announcement. That's about $100 billion in capitalization lost.

There's no inherent predilection to censor based on political ideology. It's an inmate human failing. The BJP is right wing, as is Putin. China is nominally Communist but is actually fascist, achieving political superposition and being the poster child of the independence of totalitarianism from ideology.

What Huang says is reasonable. Everyone has a different definition of AGI. If the definition is easier to achieve, like a score on a test, then five years is entirely reasonable, but many will not accept that definition of AGI. The other part of what Huang is saying (perhaps in an unspoken way) is that for other more ambitious definitions of AGI, we are quite a bit more than five years away.

All dictatorships are "benevolent." In both China and Saudi Arabia, citizens live well and are happy as long as they do "the right thing." Consider the fate of Jamal Khashoggi, who didn't do "the right thing." Also, in both countries, there are a lot of poor people who don't live well. In China, those poor are citizens, while in Saudi Arabia, they are imported non-citizens. In both countries, these poor people will not live so well, even if they do "the right thing."

We, the home enthusiast, welcome trading off our time in exchange for paying less. Of course, the home enthusiast isn't buying any of these new Nvidia systems. That's left for big companies, especially the hyperscalers that purchase half of the Nvidia data center GPUs. For the hyperscalers, time is more important than money or power. This is especially true in this wild west era of emerging models. Time to market is everything, which is why the new, more expensive Nvidia systems will sell just fine. The models are getting bigger, and R&D iterations of model improvement are happening more frequently.