stri8ted

You seem to assume training on pelican would not result in improved performance on other similar tasks. Why?

Still has a 1.78T market-cap. It will be OK.

Its actually useful, when you are launching into a long monologue and want periodic acknowledgement that its "listening".

Way too expensive. Google vision OCR (which they failed to compare against), is $1.50 per 1k pages. Vs $4 from Mistral.

What metrics are you using to classify it as a downfall?

Is there something in the post that you find implausible or don't believe to be true?

I doubt this is representative of real world usage. There is a difference between a few turns on a web chatbot, vs many-turn cli usage on a real project.

Entire segments of the podcast sphere are making their money talking about these so-called unspeakable subjects. Why don't you share what you really think.

Those are the same thing

48 GB is not consumer hardware. But fundamentally, there are economies of scale due to batching, power distribution, better utilization etc.., that means data center tokens will be cheaper. Also, as the cost of training (frontier) models increases, it's not clear the Chinese companies will continue open sourcing them. Notice for example, that Qwen-Max is not open source.

Do you have any evidence to support this view?

Price Input: $2.50 / 1M tokens Cached input: $0.25 / 1M tokens Output: $15.00 / 1M tokens

https://openai.com/api/pricing/

Flock has put out a report claiming 10% crime in the US is solved using their technology. There are of course counter argument, that claim this is not valid.

https://www.flocksafety.com/customers/how-many-crimes-do-aut...

I take into account publicly available information (news articles), factor in personal anecdotes, and reason about human nature and incentives. I know the extent of reported abuses, and I do my best to extrapolate. It's not perfect, but such is life.

To be clear, even if we all agreed on the data, I still would not expect everyone to take the same position. There are subjective differences in values.

It's clearly true there have been abuses as a result of this technology. And its also clearly true criminals have been caught as a result of the cams, that otherwise would not have been.

If you believe the costs of the the abuses, and potential abuses, exceed the benefit, then at least be honest about the trade-off, because there are real benefits.

Personally, I believe the costs, on net, are worth the benefits. And in so far as the costs can be further reduced, without loosing most benefits, then great. This is not right or wrong. It's just a question of values, and how you weight the costs vs benefits.

Don't down-vote this all at once.

Can you show some comparisons for WER and other ASR models? Especially for non english.

How is this content related to HN? Are there any submission criteria?

Exactly. As far as I'm concerned, the benchmark is useless. It's way too easy and rewarding to train on it.

That is where the money is.

What languages does it support? I can't find this info anywhere on the page.

For video use cases, which will become increasingly popular, we are a long ways away.

Exactly. Or use the interpretability work to disable the distress neuron.

The same is true for Google.

It seems a strong base model is what enabled this. The models needs to be smart enough to get it right at least some times.

Not for coding, it seems. https://aider.chat/docs/leaderboards/

Given Israel's successful precision targeting of various senior Hezb members in recent months, I wonder if the pagers were initially used as such, but as suspicion mounted, and chances of an overhaul increased, they decided to hit the kill switch while they still could.

Although as as per an WSJ article: "The affected pagers were from a new shipment that the group received in recent days"

This is not the case. From Palmer's Twitter:

"Pulsar is one of several Anduril products that has been undercover for years. It has already been deployed to multiple AORs across continents at fixed sites, on ground vehicles, and in aircraft.

RFML (Radio Frequency Machine Learning) is a game-changer for electronic warfare."

What is the pricing? Also, why the strange "neuron" pricing for CF models?

did you know the nordic countries are among the happiest and have some of the strongest social benefits in the world while still being wealthy?

I do wonder how much of this is due to cultural homogeneity and the benefits that follow. Will this remain the case in the presence of a larger proportion of immigrants? I don't think its a fair comparison, when people contrast Nordic countries with the US.

Many studies have found that social trust is higher and easier to maintain in homogeneous societies.

As far as I understand, these multi-modal models work by embedding the text/image in a shared representation space. To perform OCR on such an embedding, it would require extracting every letter, in the correct order, from the embedding. But given the embedding is a fixed size, and therefor necessarily compressed, I would expect it to loose the exactness of the underlying input, especially with images containing a lot of text. So assuming GPT-V can effectively perform OCR, how is this being done given the constraints?

Or is my understanding completely off? Perhaps it's "Translating" the image to text, by outputting a sequence of text tokens as it scans the image regions, and then the text queries (e.g. "whats funny about this") uses this translation as the context? Presumably, this is how the model handles audio input.