It’s Getting Better!
This morning, I updated my “About” page to include a new summary of my background based on a question I asked Grok 2.0-Mini. This is a new release of Grok to the general public, and I was astounded by both the level of detail and accuracy in the summary it created of my background. It made me sound more successful than I think of myself, but it’s factually correct. I read it several times to make sure it had the right context.
As I’ve said many times, I regard our success as an ensemble event, and we almost didn’t make it. There was no way we’d have achieved any of it without the technical brilliance of Greg Burns as CTO. My job was to push the company and make sure we survived an exit at the right time. Greg wasn’t just brilliant; he also ran an extremely loyal engineering team who loved working with him, and they were all amazing people whom I will always hold in the highest regard. It was Akemi Sagawa, who founded the company, Rick Romawotski as our COO, and all the others in all departments who got us there. I was not a one-man show, but I sure had to push through some stuff as CEO. It wasn’t always a fun time.
The point isn’t to write about OINA but to point out the accuracy of X.ai’s Grok 2.0 as something even more factually accurate than GPT-4o, and it was created in far less time. I’d call X.ai’s progress a monkey stomping. Right now, the only advantage that GPT-4o holds for me is that it remembers all my work. Grok 3.0 and subsequent releases will probably leapfrog that small feature.
Grok 2.0 is a very usable AI right now, and it will become a primary research tool. I have no idea where it pulled such accurate data about my background and put it together so well. I could have kept the word “Bluetooth” out, and it would have written more about my writings and Moose Lodge. So then I wonder, if it pulls from what it already wrote about me, does it allow someone to create a completely fake background? I suppose it does, but the cool thing, the reassuring thing, is that everything it wrote is verifiable with just a little bit of digging.
There is a lot that I’ve done since OINA and yet, much of that has to remain under NDA. The results are more public, but the actual work still remains a secret. Someday, I think these LLMs will triangulate and draw their own conclusions about our successes and failures. That will be an interesting time. Imaging the BS that will get flagged.