Can speech-to-text models keep up with "Breathless" by Shankar Mahadevan?
Shankar Mahadevan sang it like breathing was optional.
We put three speech-to-text models to the test. Two ran out of breath. Watch till the end to see which one didn't.
But how does Pulse do on the audio voice agents handle every day?
On real Hindi contact center calls, Pulse had a 10.20% word error rate, lower than Sarvam Saaras v4 and Deepgram Nova 3.
Full benchmark results are in the first reply.
@mitchellh It’s gotten better but not as much as everyone hoped. Especially the random delays in answering simple questions that other AI’s can answer easily are extremely irritating.
#Apple has had many product launches with controversial design decisions. They were always good at explaining the value prop, and thought process behind it in their marketing.
https://t.co/ezaPtusWrN
1/n
#MacBookPro with touch screen is the first time they have not only went against their stated principles (laid out in 2010 keynote by Steve Jobs), and while doing so, have made zero effort to explain why touch screen makes sense now. 2/n
When @typesafeai AI released #Jev 1.14 with experimental image support, I had to test it on SOS Draw!
Can @typesafeai recognize a player's raw canvas drawing better than our handcrafted stroke-to-text pipeline?
I ran a side-by-side benchmark: Text vs Image vs Hybrid. 🧵👇
I think I found a planet nobody knew existed.
It's been hiding in NASA telescope data for seven years.
At the end of this thread, you can fly to it yourself.
Duolingo should offer a way to block / pause certain characters. E.g. Lily’s nonchalant tone was funny for a few months but gets monotonous over time to the point that I skip any skippable Lily content that I can.