One thing I don't understand is google has so much metadata on search sessions to RLHF their search results.
E.g. when I start a search session to solve a programming problem (before llms), I will continually search different terms to get to my solution webpage. Then stop. This session metadata and the path I took is highly significant data that can be used to help llms recognise what research itself looks like.
The defining metric of progress in a society is that all of us have to work less for maintaining same or higher quality of life. Leaving aside the supply demand aspects, who is going to pocket the savings if people aren't overpaid and why should they be the appropriate recipients of that savings.
Building personal assistant could be beneficial to Mozilla based on how much we do online. I would like to track changes to my beliefs based on how I came across new information. In future, the AI could automatically shorten paragraphs in essays about topics or terms I am already aware of while keeping new concepts introduced in it full expanded so that I grok them better.
I think the talk is meant for executives who are more cautious and want the fluff and dust to settle before they deploy resources at their companies at the next shiny thing.
AI resides inbetween the mobile layer [hardware devices through which we will consume content] and above internet layer [information / content]. In that sense AI will eat stuff at the layer, tasks which were generally done differently.
According to Sachs, Israel has masterfully manipulated US influence to extend its global reach, primarily through AIPAC's incredibly efficient lobbying - spending just hundreds of millions to secure billions in aid and trillions in military spending. Netanyahu's strategy has been particularly clever, pushing the US to overthrow Middle Eastern governments that oppose Israeli policies, as seen with Iraq, Syria, and Libya. Through campaign financing, Israel has basically bought out Congress for surprisingly little money, ensuring the US consistently backs them internationally - like vetoing UN resolutions that favor Palestinians. This US shield is so strong that when the UN voted on Palestinian self-determination, only the US, Israel, and a couple other countries opposed it. Even when Biden sets boundaries for Israeli actions, they just ignore them without consequences. The whole system's genius lies in how Israel's managed to maintain its policies despite global opposition, though Sachs thinks this might backfire by making Israel too isolated and blocking any chance of a two-state solution.
Is there an example of a company which could have gotten a new lease of life after acquition was denied a regulatory approval and then the company failed?
Meta started carrying out the last batch of a three-part round of layoffs on Wednesday, according to a source familiar with the matter, as part of a plan announced in March to eliminate 10,000 roles.
The media was lying all the time. It didn't start lying just after advent of social media. Just look at sitcoms like Yes Minister. The reason we now consider media to lie is because their lies are exposed, documented, tracked and refreshed to remind users not to trust them blindly.
Nobody will visit stackoverflow because AI through its reasoning and back and forth with users will have solved the problems. This process creates training data for future AIs of that particular company unavailable to any other
Second issue: who decides the weights of sources. this is the reason why every nation must have culturally aligned AIs defending their ways of living in the information sphere.
I have a physics analogy which is similar. Vested interests set up magnetic fields in social media / legacy media (lot of things discussed prominently in social media is just what legacy media is saying. So legacy media is a sense is setting up an anchor points and people have to distribute themselves around it) to flip magnetic domains to align with the narrative.
This is true of the feeds and everything else where there are abundant choices. Amazon putting its inhouse brands before others. Anything which has to be narrowed down is an algorithmic choice, either data driven or top-down.
I can suggest one reason why LLM might prefer writing in higher level language like Ruby vs assembly. The reason is the same as why physicists and mathematicians like to work with complex numbers using "i" instead of explicit calculation over 4 real numbers. Using "i" allows us to abstract out and forget the trivial details. "i" allows us to compress ideas better. Compression allows for better prediction.
In a large orginaization, launching a new feature requires interaction with lot of other existing teams which slows speed down. Auth, permissions, metrics, etc. The new feature must seamlessly interact and integrate with existing systems.
I have a offtopic but related question. Why are our taxes still calculated in stepwise bracket manner. Why not a smooth curve? We have the mathematical and technical knowhow to implement it. Coming back to the Dropbox layoffs, why was this decision a drastic stepwise reduction. Not a smooth curve over a period of say 1 year.