Now I get it why Trump saved Tiktok :)

If you want to learn more about how this works, be sure to read on!
www.emsi.me/emsi/article...

Scientists have shown that even smaller models, like those with just a few billion parameters, can achieve results comparable to the largest AI systems. How? Through an intelligent approach, combining code with logical reasoning and the ability to self-correct.

What’s important is the habit of noting down important thing so it’s easier to reference that later without spending too much time on finding where it was.
OneNote works for me as it can search though all of the notes I’ve ever made.

When I make notes (and I do not always make notes when reading technical book, only when I’m reading something new that is not easy to understand or something that I find… erm… noteworthy) I learn. There are notes that I returned to many times and there are notes that I never had to return to.

www.youtube.com/watch?v=F_Ri...

For those working in tech: How do you separate genuine technological breakthroughs from marketing in emerging fields? What metrics do you look for?
#QuantumComputing #ComputerScience #Technology #Innovation #TechnicalDiscussion

Looking at Google's announcement through this lens helps understand both its significance and limitations. While we're still far from practical quantum supremacy or breaking cryptography, seeing improved coherence in circuits with increasing complexity is noteworthy for the field's development.

It brilliantly explains quantum computation through basic linear algebra - qubits as two-dimensional vectors, quantum gates as matrix operations, and how quantum phenomena like superposition and entanglement translate to computational advantage. I'll share the link in the comments.

This brings me back to when I first started understanding quantum computing beyond the usual "it tries all possibilities simultaneously". If you're curious about the actual computer science behind quantum computing, I highly recommend Andrew Helwer's talk "Quantum Computing for Computer Scientists."

While we should be cautious about extrapolating this pattern to much larger scales, it's an intriguing development in addressing quantum decoherence, one of the field's fundamental challenges.

The team demonstrated a computation that took their quantum chip about 5 minutes, while theoretically requiring around 10^25 years. What fascinates me most isn't just the speed comparison, but the reported improvement in error rates as the circuit complexity increased - from 9 to 49 gates.

This approach mirrors human cognition, where not all reasoning is verbalized. COCONUT uses special tokens to manage this latent reasoning mode, leading to impressive results: 34.1% accuracy on GSM8k math tasks and 97% on complex planning tasks, outperforming traditional methods.