HN user
I don't understand:
all of the prompt tokens are available in advance and do not require decoding
The other technique is called batching: this technique aggregates multiple prompts into a single decode operation.
So do prompts get decoded or not? Are there 2 decode steps? Unclear
This would lead to a rise in doctors and accountants wearing the combo of Canada goose jacket, balaclava and man-sling-bag in summer.
This is untrue? Cameras in the UK are not "just used for surveillance". The facial recognition used recently has led to arrest of many offenders.
The Met reported that in 12 months they made 580 arrests using LFR for offences including, rape, domestic abuse, knife crime, GBH and robbery, including 52 registered sex offenders arrested for breaching their conditions.
https://www.gov.uk/government/news/live-facial-recognition-t...
That's statistics for London, not the rest of the UK.
I thought it was pretty rhythmic
I would've believed he's real, just passionate about music on his big yellow bus.
The only time I think about using or paying for Kagi is when I see these posts on HN and get FOMO. I think that's because I actually don't have that SEO garbage problem anymore. AI apps have made search, and learning in particular, a lot easier.
Is this article roughly the application of the startup mindset to research? With the major downsides that you're giving away your insights for free, and potentially actively help other companies profit from your work. You don't go as far as to productize the research, but just make it really easy for others to do so by building "artifacts" for it.
I'm not a researcher, but have thought about doing a PhD in the past.
It's probably a lot more nuanced that this. Show progress but don't make it easily accessible. *Hide* something important for yourself. Kind of like modern day "open source".
Yes we have to make users happy. But my mission is to build stuff so that bad stuff doesn't happen: Russia doesn't invade Ukraine or China doesn't invade Taiwan.
It depends on the use case. Some of these startups and unicorns ship (and aim to ship) stuff to Ukraine. That's one conflict most people in the west can align with. Imagine building drones or software for Ukraine. Other use cases involve purely intelligence or decision making.
Sure they're building stuff for Ukraine soldiers, but the underlying mission is freedom for Ukraine.
There are other conflicts (building drones, missiles, tools used in Gaza/Middle East) that are a lot more complex.
Or building drones for police pursuits or for searching buildings (like the IDF).
The whole concept is pretty stupid, rooted in a silly assumption; that a grand, noble mission is essential to motivate employees.
I am motivated by an important mission, in fact, but I am more motivated by having ongoing impact on that mission (making a difference). I work in defence, so you'd think there's a lot of opportunity to keep people safe, help leaders make better decisions, etc. So far from my perspective, these companies are not very motivated or successful at solving the missions they boldly claim. As a result, I am motivated to leave these places, that lie or are incompetent at solving these missions.
For example, a defence company says they want to "serve our democracies", but they hardly have any users. Another defence company has been building product for 4 years, yet has 0 users over that time. Then there are other, defence companies with unicorn valuation with also very cool marketing, but very little impact in ongoing crisis. All the mentioned companies have very good publicity, it's not until you work for them you realise they have 0 impact. There is no reason to change their ways, because they can continue to sign more contracts with gov (tens of millions, hundreds of millions).
All you see publicly is cool tens/hundreds of million investment or contracts, cool movie like drone shots, "AI", cool drones, explosions, jets and good music. You don't get to see the 0 impact. Or worryingly, negative impact when these things don't work.
I've been working in defence-tech startups for more than 2 years.
Wasn't he was fired or asked to leave from DeepMind, then Google?
I interviewed at Inflection (his startup), and my interview with him (final stage, founder interview) was the worst interview I've ever had. He did no small talk, he immediately (and awkwardly) asked a question off his question list. It lasted 15 minutes. I was immediately put off by his questions, wondering if I should just hang up now. His co-founder interview was not as bad, but still asked an awkward question. They definitely want you to work very hard, including sacrificing your personal life.
The product itself (Pi, the assistant) doesn't seem to have traction. Has anyone found it useful?