Browsed by
Category: AI

AMD accidently releases FSR

AMD accidently releases FSR

Comments 0 Comment

AMD recently published a new version of its FidelitySDK with FSR 4 upscaling and FSR 3.1.5 frame generation support. Unfortunately, they accidentally published the full FSR 4 source code on GitHub. Before AMD took it down, some media outlets and X users managed to record screenshots of the files, including Videocardz.

Update: PCGamer is reporting that AMD is having more trouble putting the genie back in the bottle than first thought because they accidently included the MIT license.

Which AI’s are best?

Which AI’s are best?

Comments 0 Comment

Retired programmer Dave’s Garage decided to look into all the major LLM AI models and gives his feedback on using them for the last few months.

Which should you use? It depends on what you’re trying to do. It also depends on how you’re testing it – because others come up with different ratings.

Coding

  • Claude 3.7/4 is great for serious/production oriented code (after a code review). Probably the winner here.
  • ChatGPT 4.1 is a good copilot for prototyping and exploration
  • Grok 3 – responds quickly/good speed
  • Gemini 2.5 Pro – does what you ask but not much more

Research

  • Claude 3.7/4 for carefully explained reasoning and good for references
  • ChatGPT 4.1for clear overviews
  • Grok 3 for current event
  • Gemini 2.5 Profor large, structured input and extraction

Storytelling

  • Claude 3.7/4 – literary and reflective
  • ChatGPT 4.1- most emotionally resonate
  • Grok 3 – flexible and imaginative
  • Gemini 2.5 Pro- informative and expandable

News

  • Grock 3 wins this easily – gives news and what people are saying about it
  • ChatGPT 4.1- Can handle current events decently but slower to pick up news
  • Claude 3.7/4- largely sits out news and doesn’t comment unless widely verified
  • Gemini 2.5 Pro- factual, accurate, but rarely first

He also discusses the different context sizes when it relates to the tasks. Bigger windows cost more but can allow you to summarize huge codebases or 60 page complex legal documents.

ChatGPT can handle 128,000 tokens or about 96,000 words (1 token roughly equals 4 characters). Claude has 200,000 tokens or about 150,000 words. Gemini 2.5 Pro and Grock 3 claim to have 1 million tokens.

If all you’re doing is summarizing emails, ChatGPT could be just fine. But if you need to make sense of large codebases or summarize large legal briefs, Gemini or Grock will be better at avoiding hallucinations or leaving gaps. There are some that believe that these windows might actually shrink if the systems are under heavy load (Grock in particular).

AI written mushroom foraging books will kill you

AI written mushroom foraging books will kill you

Comments 0 Comment

Atomic Shrimp noticed that a number of recent mushroom foraging books had errors. These weren’t simple errors, if you ate some of the things they say you could eat, you could destroy your kidneys or even kill yourself. How did this happen? He realized a lot of these books were being generated by AI and it looks very much like the people that put them together didn’t even know how to fact check them.

It’s a great discussion of how AI generated books have become very prevalent and the dangers of people just churning out AI slop without quality control.

25% of all new code at Google is AI generated

25% of all new code at Google is AI generated

Comments 0 Comment

Google’s CEO revealed that AI systems now generate more than a quarter of new code for its products, with human programmers overseeing the computer-generated contributions. The statement, made during Google’s Q3 2024 earnings call, shows how AI tools are already having a sizable impact on software development.

Stack Overflow’s 2024 Developer Survey, over 76 percent of all respondents “are using or are planning to use AI tools in their development process this year,” with 62 percent actively using them. A 2023 GitHub survey found that 92 percent of US-based software developers are “already using AI coding tools both in and outside of work.”

https://arstechnica.com/ai/2024/10/google-ceo-says-over-25-of-new-google-code-is-generated-by-ai

AI shows you the faces of ‘average’ British cheaters

AI shows you the faces of ‘average’ British cheaters

Comments 0 Comment

2,000 Brits recently took part in a study by online casino MrQ to see if AI could make a picture of the ‘average’ British cheater. They collected descriptions and then asked AI to make a composite image.

The average male cheater that seemed to be someone in his forties, with blue-grey eyes, small lips, short facial hair and little to no head hair. Someone who has a larger nose and visible frown lines.

The typical woman who cheats, apparently, has dark-haired and is in her early fifties. AI reckons that they are more likely to have a small nose and a medium-sized pout.

Links: