Nathaniel Whittemore
speaker
52,364 appearances
204 recordings
3 series
first heard Oct 2025
last heard yesterday
Nathaniel Whittemore’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 204 in all, peaking in Jun 2026 with 31.
Appearances
Yesterday, Diogo Almeida posted on X, After co-inventing ChatGPT, I kept asking myself, why have superhuman chat models not led to AGI?
podcast-host
I've spent the last two years in stealth building a new way to train models, RLCD, or reinforcement learning for calibrated decisions, and a new type of frontier AI model that we are releasing today.
podcast-host
Jeff.
podcast-host
JEV is 20-200 times faster, 40-400 times cheaper, with output tokens free, frontier composable intelligence optimized for decisions.
podcast-host
As far as I can tell, the shortest path to AI-based economic revolution.
podcast-host
Now, one could absolutely be forgiven for seeing numbers like 20 to 200 times faster and 40 to 400 times cheaper, and being a bit skeptical of the claims to say the least.
podcast-host
And were this just another LLM, that skepticism would be entirely warranted.
podcast-host
However, with Jev and this new strategy, we're dealing with something that is quite different.
podcast-host
The company behind Jev is called TypeSafe, and in the announcement blog post, they talk a little bit more about what makes their approach different.
podcast-host
Right, we built a new stack entirely focused on automation, with a new model architecture, parallel sampler for maximum efficiency, and training method we call reinforcement learning for calibrated decisions.
podcast-host
Whereas existing LLMs optimize for human preference, i.e., write-ups and chat responses that human raiders prefer, the new System 1 models, the first of which is JEV, optimize for calibrated decisions, or answers with epistemically honest probabilities.
podcast-host
Yeah.
podcast-host
So what does that actually mean?
podcast-host
Well, let's look at how Mike Taylor from Every describes it.
podcast-host
He writes, think of it as a smart if-then statement that determines what happens next when you're automating a workflow.
podcast-host
Say you're building software that prioritizes customer service requests, and you write code that asks the model, does this customer sound angry?
podcast-host
Jev might answer 0.9, which means there's an estimated 90% probability that the answer is yes based on.
podcast-host
What the model learned in training.
podcast-host
You could also provide categories you define like annoyed, irritated, offended, furious, and enraged, and learn that the customer was 60% likely to be classified as furious, with only a 10% probability of being enraged.
podcast-host
Going on to explain why this matters and how it differs in LLMs, Mike continues, with an answer of 0.9, very likely to be angry, the software might automatically proceed to escalate the
podcast-host
Showing 3041–3060 of 52,364 · page 153 of 2619
← Previous
Next →