The link also breaks if the original user deletes the chat that was being linked to, whether on purpose or without realizing it would also break the link.
In my experience, iOS 18 makes it better, but no, it is still not as good as Whisper Large, which is what OpenAI uses behind the scenes.
You can run smaller Whisper models on your phone through various third party apps, and those options work well too. (Trying to run Whisper Large even on the iPhone 15 Pro Max is asking too much… it works, but only barely.)
Immediate example that comes to mind: there’s a paid-for Canva account that multiple people need to use. Can’t use separate logins because then you’d need multiple subscriptions.
Moral argument is that a PTA are part time organizations and if canva doesn't cater to that business model, then shared credentials is fair use, commercial and for profit use is a horse of a different color.
It’s weird how non-specific the CPU is there. Why wouldn’t they list a CPU part number? We don’t even know what generation of Epyc it is. (I get that it’s not the focus… but it is still important.)
The Pi5 southbridge was codenamed RP1, the RP2040 was RP2, and the RP2350 is apparently RP4. What might RP3 be I wonder? The boring prediction is that it's just a new southbridge for the Pi6, but one can hope they eventually split with Broadcom and do their own in-house Linux SoC which they would be able to document properly.
This might help explain in-part the discrepancy. Perplexity.ai[1] also says 8,200 German postal codes. I set Claude 3.5 Sonnet in the LLM settings on Perplexity but it looks like it might use a ChatGPT model for the initial search of sources? At least we can see what it is sourcing to fetch the value of 8,200. Interestingly, asking Claude 3.5 Sonnet directly at claude.ai returned 16,000.[2]
1. The first digit represents one of 10 postal regions.
2. The second digit typically represents a sub-region within that area.
3. The last three digits identify specific delivery areas or post offices.
It's worth noting that the exact number can fluctuate slightly over time due to administrative changes, urban development, or postal service reorganization. However, 16,000 is a good approximation for the total number of German postal codes.
Would you like more information about how the German postal code system works or its history?"
Llama3.1-405B benchmarks at a level that I would certainly consider comparable to other SOTA models, as long as multimodality isn’t needed.
Regardless, OpenAI provides access to quite a few of their older models through the API, since the API lets you pass in a specific model version. I’m sure the older models won’t be available forever, but that is a much more stable target for researchers than just opening the ChatGPT website and typing in things.
The Google Authenticator app (just as a mainstream example) was released 14 years ago. When we're still waiting for a lot of banks to even support TOTP, consider me unimpressed with the level of effort banks are putting into securing my accounts.
