pull down to refresh

From @Space_Child67:

As I understand it, quantum computing is helpful for solving NP-hard problems, and NVIDIA has introduced QUDA to enable hybrid quantum-classical computing with applications spanning drug discovery, chemistry, weather, finance, logistics, and more. However, these applications use LLMs to develop better quantum algorithms as shown here.

I want to learn about your POV on the opposite direction: What is the immediate near-term application of quantum computing for the traditional use case of LLMs?Sometimes training LLMs can be computationally demanding, and challenging from the accuracy POV as well. Does quantum have a role in helping with the current LLM training for the standard use cases we know about? Here is an article from IONQ. Thanks in advance for your response.

Sorry to burst the bubble, but it's not obvious to me whether quantum computers will help at all with training LLMs. Maybe eventually the Grover speedup will help somewhat. Or maybe there are new quantum algorithms yet to be discovered that will help, or old quantum algorithms that will turn out to work really well when we test them at scale on AI problems. Sadly, though, the articles you'll find on "quantum AI" tend to contain ENORMOUS amounts of misrepresentation and hype -- often simply failing to compare to the best classical solutions that are available, and hoping no one will notice that.

reply