Garrison Lovely

speaker

631 appearances 3 recordings 1 series first heard Jul 2026 last heard 15 Sep

Garrison Lovely’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 3 in all, peaking in Sep 2026 with 2.

Appearances

newest first · ▶ plays the moment

Breaking Points with Krystal and Saagar · 9/4/26: Diesel Prices Skyrocket, Jeff Sachs Warns Of Tech Collapse, OpenAI Hack Coverup, JD Responds To Tucker Praise Of Abdul, Fox Fires Maria Bartiromo · 4 Sep 2026 podcast

It didn't always include everything the model was thinking, um, but it was way better than nothing.

And it was like this incredibly lucky break.

And then, you know, earlier, what no last week now, I guess, um, or earlier this week, God, who can keep track of?

There's this report in the information that OpenAI has started using this technique, which

basically lets the model think more in its head before it does a next step that it writes down.

And this is something that like people have known about for a really long time.

And it's very concerning because it's like kind of obvious why this would be helpful to make the model more capable, more efficient, and then also why it would be like dangerous to have the model be able to do that uh without you seeing as much of its, you know, thinking.

And you can put air quotes in, you know, around the word thinking for all of this, but just like that's a helpful way to understand like what the model is doing here.

And

It's, you know, some open AI safety people who I think are actually credible, despite working at a company that is deeply uncredible, um, have said like this new breakthrough technique is not the reason that Astra GPT six is so much less understandable.

Um, which like, okay, let's assume that's true.

It's like that's also concerning.

You just don't know why.

Um and so either way, it's like the the model

can do just so much more in a single forward pass uh of of like when it's prompted, um, without having to do these intermediate steps, where it was like going from like a few minutes to like 30 minutes on like a math problem, like in terms of like how long it would take, you know, a a person to do it.

And this is obviously really concerning.

And the the the breakthrough, like when I saw the benchmark scores for GPT six,

I thought they were fake.

I didn't recognize the source, but they were just so ludicrously high that I'm like, surely this can't be real.

Um and I had a feeling of vertigo.

Showing 401–420 of 631 · page 21 of 32