HN user
Hard to take Gary seriously calling out "outlandish claims" with "no substance" when he does the same thing in the opposite direction.
Garbage article for clicks to pay for his lifestyle, now that he's grifted his way into being an "AI Expert" paid to pontificate with no skin in the game.
The section 174 changes were never intended to go into effect and are a byproduct of the bill review process by deficit hawks. A group in the House has called for changes but a bill hasn't made it out of committee yet.
Actually, this has the most impact on startups that are close to break even or barely making profit. Let's say you make 1m ARR and have 5 software engineers at 200k/year. Prior to this tax code, you had $0 in profit to be taxed. With this change, you can only amortize 10% of software labor in the first year. Now, the IRS treats you as having 900k profit so you have ~200k in taxes to pay.
Imagine you're a small business that makes $1m ARR and you employ 5 software engineers at $200k per year. Your net income is 0.
Prior to this change in the tax code, your software labor costs for that year would all count against your income so you'd be taxed on $0 in profit.
With this tax code, you can only amortize 10% of software labor in the first year so now your business just had $900k in profit as far as the IRS is concerned. You now have to pay ~200k in taxes. You have to come up with that money somehow, and for most businesses the only short-term option is a combo of reducing costs (layoffs) and loans.
It’s “underused” precisely because this feature is pretty much useless if you’re transcribing anything other than quick snippets of speech.
I'm not sure why you're so dismissive when real-time transcription is an important use-case that falls under that bucket of "quick snippets".
It’s also hard to use since you have to know in advance what hard-to-transcribe words are going to be in the audio.
I think it's more context-dependent than it is "hard". It's ideal for streaming meeting transcripts. In my use-cases, I use the prompt to feed in participant names, company terms/names, and other potential words. It's also much easier to just rattle off a list of potential words that you know are going to be in the transcription that are difficult or spelled differently.
We need a better solution. It would be much better if there were an easy way to fine tune Whisper to learn new vocab.
Prompting is infinitely easier than fine-tuning in every aspect. I can reuse the same model in any context and just swap out the prompt. I don't have to spend time/money finetuning... I don't have to store multiple fine-tuned copies of whisper for different contexts... I'm not sure what better solution you envision but fine-tuning is certainly not easier than prompting.
Microsoft has the most well-regarded security reputation amongst all cloud providers, it's the main reason why they were awarded the (since canceled) JEDI defense contract.