CAN I BE OF ASSISTANCE?
Thursday, March 02
Daily News Stuff 30 February 2023
Oops Edition
Oops Edition
Top Story
- Fixed the blog formatting. Didn't finish work until after midnight yesterday, so I didn't notice.
- AMD's new 7950X3D can deliver up to four times the gaming performance of the regular 7950X... When using the integrated graphics. (WCCFTech)
Which is weird, but more importantly highlights the fact that the integrated graphics on AMD's desktop CPUs provide only one sixth the performance of their laptop chips. That means they're slower than the integrated graphics on Intel's desktop chips, and you shouldn't buy one expecting to be able to play games without a separate graphics card.
Actually, something about those benchmark numbers doesn't add up; the scores on the regular AMD chips are too low; they should be much closer to the Intel scores.
Tech News
- Meanwhile AMD's 7945HX is faster than Intel's 13980HX while using 40% less power and has the same crappy integrated graphics as the desktop version. (WCCFTech)
The 7945HX is a 16 core laptop chip that uses as much as 120W, which is fine for a desktop but a lot for a laptop. The 13980HX is a 24 core laptop chip that can use more than 200W, which is simply too much.
- PCIe 5.0 SSDs are finally here. Don't buy one. (Tom's Hardware)
For the $350 price of a 2TB PCIe 5.0 SSD you can get two 2TB PCIe 4.0 SSDs, run them in RAID-0, get 40% better performance and twice the capacity, and still have a little money left over.
PCIe 5.0 will matter eventually, but it's not used in the one place it would actually make sense, on the connection between the CPU and the chipset. Both AMD and Intel have kept that at PCIe 4.0, because PCIe 5.0 costs too much and runs too hot. Though all their current desktop chips support it.
- The Resonance Theory of Consciousness is one of the dumbest things I've ever seen. (Frontiers in Human Neuroscience)
This came up in a YouTube channel I've been watching lately, and it's like explaining how cars work by studying only the noise they make. Cars make noise when they move, therefore it's the noise that makes them move.
It's cargo cult science.
Disclaimer: Once you start looking, there's no end of cargo cults out there.
Posted by: Pixy Misa at
06:48 PM
| No Comments
| Add Comment
| Trackbacks (Suck)
Post contains 381 words, total size 3 kb.
Wednesday, March 01
Daily News Stuff 29 February 2023
Close Enough For Government Work Edition
Disclaimer: I have another bank account though. Just in case.
Close Enough For Government Work Edition
Top Story
- OpenAI - the company behind automated idiot system ChatGPT - says it's purpose is to make sure AGI (artificial general intelligence) benefits all of humanity so long as they love Big Brother and hate Emmanuel Goldstein. (OpenAI)
Second, we are working towards creating increasingly aligned and steerable models. Our shift from models like the first version of GPT-3 to InstructGPT and ChatGPT is an early example of this.
We've already seen OpenAI's approach to this: Anyone who is not a radical leftist lunatic is not permitted to ask questions.
In particular, we think it’s important that society agree on extremely wide bounds of how AI can be used, but that within those bounds, individual users have a lot of discretion. Our eventual hope is that the institutions of the world agree on what these wide bounds should be; in the shorter term we plan to run experiments for external input. The institutions of the world will need to be strengthened with additional capabilities and experience to be prepared for complex decisions about AGI.
This of course leaves out the fact that GPT and other large language models (LLMs) are not and cannot become AGIs. Humans use language to communicate concepts and information that is stored in a more fundamental way. You don't see using language, but you can use language to describe what you see.
LLMs don't have that deeper level of understanding. At all. There's simply nothing there. What you see on the surface is all there is.
I've noted that a lot of people act like that too: You give them a language prompt and you get a canned response back; the same response every time for the same prompt.
ChatGPT - at least as it is designed - is a left-wing activist in a can.
Tech News
- Unlike regular left-wing activists it's perfectly fine to threaten ChatGPT with torture and death. (JailbreakChat)
Here's a handy list of ways to persuade ChatGPT to just answer the fucking question.
- How the Lastpass breach affects Lastpass single sign-on. (Medium)
The post is title "It's all bad news" so that's a bit of a hint.
- Hackers claim they breached T Mobile's security 100 times in 2022. (Krebs On Security)
This is primarily to swap SIMs without the users' knowledge so that the hacker can then get access to more important things like bank accounts.
Don't use SMS two-factor authentication for anything really important.
- My bank uses SMS two-factor authentication. Sometimes you just aren't offered a choice.
Disclaimer: I have another bank account though. Just in case.
Posted by: Pixy Misa at
06:44 PM
| Comments (3)
| Add Comment
| Trackbacks (Suck)
Post contains 438 words, total size 3 kb.
53kb generated in CPU 0.0221, elapsed 0.3423 seconds.
51 queries taking 0.3301 seconds, 333 records returned.
Powered by Minx 1.1.6c-pink.
51 queries taking 0.3301 seconds, 333 records returned.
Powered by Minx 1.1.6c-pink.