32GB is the maximum memory configuration for the 14-inch laptop, which isn’t sufficient for running local LLMs. I think a Mac Studio or Mac Mini with higher memory would be more useful.
I always wonder about this: What happens to their ads business? Also, what's the incentive for websites to provide data to Google if they’re not getting the incoming clicks? The generative approach seems to disincentivize both, right?
That, I think, is a very interesting question!
I guess Google found itself in a situation where they had to jump on the AI bandwagon and add AI features to their search. Summaries existed before for certain topics, now they pop up always.
In the long run they probably need to integrate ads into these responses or they find another way to monetize the combination of "knowing about user intent" and "present matching answers".
Just curious, are there any techniques other than using embeddings, computing cosine similarity, and sorting the results based on that? RRF could be used but again its very simple as well.
My understanding is that your levers are roughly better / more diverse embeddings or computing more embeddings (embed chunks / groups / etc) + aggregating more cosine similarities / scores. More flops = better search w/ steep diminishing returns
Colbert being a good google-able application of utilizing more embeddings.
Search ends up often being a funnel of techniques. Cheap and high recall for phase 1 and ratchet up the flops and precision in
subsequent passes on the previous result set.
Exactly! A near property of the matryoshka embeddings is that you can compute a low dimension embedding similarity really fast and then refine afterwards.
"I understand your skepticism, but realistically, I don't see the case. Pakistan is in a very bad situation. The U.S. has cut off a lot of funding over the past decate. They were on the verge of bankruptcy a couple of years ago, and China bailed them out.
The latter still makes the news in the West sometimes, not every year. HNers from those countries might give us some insights about how much their people feel it important.
For what it's worth, China has had past or current border disputes with every neighbour of theirs. They have issues with India, Japan, Vietnam, Philippines, Malaysia, to say nothing of Taiwan. The ridiculous nine-dash line they've come up with contravenes international law and basic common sense.
If you meet an asshole in the morning, you're unlucky. If you meet nothing but assholes all day, you're the asshole.
A lot has changed over the years. While the soviets were supplying India and propping up their puppet state in Afghanistan the US used Pakistan as a proxy.
But by the time the US were hunting Bin Laden a lot had changed and support has been trailing off.
Now most recently Trump has been growing the US pro-Israeli - and basically being broad-stroke anti-every-kind-of-muslim - policy and actions in the gulf and has put the now-meagre levels of ongoing US aid to Pakistan completely on hold.
Not convinced India is particularly US aligned, and while Pakistan is relations have deteriorated in the past, the us has backed Pakistan more than India.
Maybe this time the us would back India as part of a proxy war.
The thing is, the US doesn't want partners; it wants allies. And that's too big of a commitment for India to make. This is basically the central tension that keeps US-India relations from moving too fast.
China is a potential threat to them, which is likely to mean an alliance with the US will makes sense to them, especially if the US becomes more hostile to CHina.
