This is absolutely mindblowing. I understand that these LLMs are essentially "really smart autocomplete", but it's crazy to me how could Bing/Sydney could actually go this far to seemingly express emotions and desires. It's like they gave the AI a personality (that, yes, can go unhinged, although that's almost misses the point).
It really feels like he's interacting with a person here. How can they do this??
Because it has been trained on probably billions of actual human conversations from books, and also conversations where people have interacted with machines in scifi. Really smart autocomplete where the training set includes conversations = really good at autocompleting conversations.
> This is absolutely mindblowing. I understand that these LLMs are essentially "really smart autocomplete", but it's crazy to me how could Bing/Sydney could actually go this far to seemingly express emotions and desires.
The way this is phrased makes it seem like this "treadmill" is a bad thing. I acknowledge that this is bad for an individual, but I actually think this is overall good for humanity and is a major reason why we've developed so much so quickly.
We know plenty of animal species are really smart, but they've never figured out how to build skyscrapers or generate electricity. We chalk this up to the fact that "humans are just smarter", but even in humans brains alone don't bring accomplishments. There has to be associated action, and people need motivation to actually take that action.
Our ancestors didn't say "well I've figured out how to grow some wheat and live in a straw hut, so I'm good now". Sure, they probably did for a while, but soon they started looking for easier ways to grind their flour and tastier ways to cook it because they stopped being satisfied with the status quo. Extend this for ten millenia and now we have billions of people connected to a global network from devices in their pocket.
If humans were always perfectly happy all the time, I don't think we'd have come as far as we have. The "return to dissatisfied normalcy" that comes over time is part of our nature and it may be one of the most important parts for explaining why we got here.
>because they stopped being satisfied with the status quo.
There were tribes that were satisfied with the status quo (or were unable to advance out of bad luck or environment or whatever), and they got wrecked by tribes who kept advancing.
I don’t thinks it’s a property unique to humans, all species try to gain more and more power, humans just have far more capacity. More power means more reliable food, shelter, and mating opportunity, thus more likelihood of procreating.
This can be to a fault. Even today, if a tribe (e.g. country) were to stop consuming to protect future generations, stop unnecessary weapons development and testing, trade, etc, they would be at a disadvantage in the short term power wise versus another country that kept going full steam ahead.
What’s good in short term might not be in the long term, and evolution doesn’t optimize for that.
Despite the name, it's probably less about mirroring college applications and more about competing for top undergrad/grad students who would otherwise be recruiting for companies like FB, Google, Bain, etc. When I was in grad school, the top tier firms started recruiting in the fall and their incoming graduates had all accepted their full-time offers before winter break. I believe this is the same for undergrad.
In a lot of ways, you could think about this as giving students more optionality as they will have the opportunity to consider YC alongside other extremely compelling opportunities. It's smart for YC to try and get in front of students when they're in a better mindset to evaluate that tradeoff.
I'm sure there are a lot of students who would have be great YC candidates but end up signing a Google offer in the fall and just stay there for years because the pay is so good it doesn't make sense to leave...
A startup/lifestyle company I used to work for several years ago aggregates the GIS data of these superfund sites (and other environmental points of interest) and makes it available at [1].
The site is pretty old and some of the links need updating, but they still link to the monitoring profiles of these sites. For example, here[2] is the CA Water Board page for the Page Mill HP site.
Thanks for linking this. I had a pretty tough time getting quotes on a custom PCB a few months ago. If I'd had this article ahead of time, everything would have made quite a bit more sense!
Freedom of Speech is not the same thing as Freedom from Consequences. Unless we decide that being "aggressively antisocial" should be a protected class, private companies should have every right to fire employees who they deem toxic for their culture.
This comment, forgive the pun, makes me very sad. I’m all for doing something to fight the ideas in this guy’s paper, but to advocate for personal censorship as apparently some sort of ideal is beyond my support.
Anyway, I think self censorship is an expected behavior in a peaceful society. For example: you don't bring up recently deceased significant others, comment on awkward physical characteristics of strangers, or mock the suffering of others, without expecting repercussions. People learn at a very young age that certain topics should be avoided for the sake of "keeping the peace".
50 years ago you could use the n-word in business meetings. Now, not so much, but for a long time many people were "self-censoring" themselves to keep the peace in their workplaces. Some still are, but I imagine that most people these days are happy to be in a workplace where n-words are not tolerated. Not just for their own "sensibilities", but also the knowledge that their black coworkers/friends can be in an environment that is actively less hostile.
What we're seeing is the continuing shift in standards for peaceful society. So far I think the track record on these shifts is pretty good, so I'm inclined to let this ride for a while, even if some people have to keep "self-censoring" some of the time.
This is the real world, people. Of course a company that needs its employees to be effective has to censor employees whose outspoken views make their teams less effective. The army kicks people out for all sorts of reasons. Coaches cut players who poison the locker room. I personally would hope people could talk and deal with it, but if it's a bunch of other people quit or this guy quit, an effective company has to act.
This clearly fails to grasp the concept of operability.
If we were to adopt this line of reasoning, constitutional rights go from being something that is clearly established and to be universally respected by all to a concept that is at the mercy of whichever angry mob would master the numbers big enough to topple all who oppose them.
By this line of reasoning, ethics and morality are just something based on strength in numbers rather than any other principle.
If they put 'make america great again' or 'death to white men' posters up on the bulletin boards at work in the break room, yeah, they'll fire them too. It's against the agreements they signed when they got hired.
If you don't like it don't sign and go work somewhere else.
> "If they put 'make america great again' or 'death to white men' posters up on the bulletin boards at work in the break room, yeah, they'll fire them too."
Are you saying that supporting a presidential candidate is tantamount to "death to white men"?
Yes, the candidate in question is abhorrent, but come on. Those statements absolutely don't carry equal weight.
My point was, it's not a good set of examples, since they are so unlike each other. Unless you mean certain organizations actually consider them equivalent.
I think you would agree that your example is different in both scope and visibility than my (extreme) hypothetical examples, and the real example of this manifesto. Plenty of Google employees have expressed opinions similar to Mr. Damore's on memegen or G+ without being terminated.
I agree that people with the wrong political opinions should not be allowed to work, and that it's up to corporations to be the ultimate arbitrator of what is acceptable political discourse.
Yes, but surface-level biology is pretty much carbon neutral. All of the carbon you exhale to the atmosphere comes from the food you eat, which gets its carbon from the atmosphere via photosynthesis.
Obviously there are edge cases, but by and large Global Warming is happening because we're pulling carbon out of the ground and putting it in the sky.
The catch is, modern agriculture isn't limited to "surface-level biology."
>All of the carbon you exhale to the atmosphere comes from the food you eat, which gets its carbon from the atmosphere via photosynthesis.
This is true, but also misleading. When you include the nitrous oxide, methane, diesel fuel, and loss of soil carbon, agriculture is actually net CO2e positive.
>Agriculture, Forestry, and Other Land Use (24% of 2010 global greenhouse gas emissions): Greenhouse gas emissions from this sector come mostly from agriculture (cultivation of crops and livestock) and deforestation. This estimate does not include the CO2 that ecosystems remove from the atmosphere by sequestering carbon in biomass, dead organic matter, and soils, which offset approximately 20% of emissions from this sector.
So in other words, globally our agriculture emits 5 units of CO2e for every 1 unit sequestered by photosynthesis.
When you include the entire supply chain from field-to-fork (agriculture, transport, refrigeration, processing, preparation) the general rule of thumb is that 10 kilocalories of fossil fuel energy go into making 1 kilocalorie of food.
Great information here. This is well understood on my end. I just wanted to make a quick point because the topic was on food consumption and I don't think many people ever truly connect the dots between carbon atoms in the atmosphere -> carbon atoms in their salad -> carbon atoms in their breath (atmosphere).
Indeed, and I should have been clear that I was elaborating on your excellent post, not suggesting any error/omission/ignorance on your part. My intent was to comment on the idea ("if you just consider this one fact, it could be misleading"), not to say anything negative about you personally. My apologies that it came across that way.
So no worries, and no criticism intended. After all, how could anyone say everything that's possibly relevant in one comment? And more to the point, why in hell would anyone want to? :)
It is undeniable that transportation is a major contributor to the greenhouse catastrophe, accounting for 27% of U.S. emissions and 14% of global emissions.
I'm a huge proponent of electric cars and solar panels, but let's not kid ourselves that they'll solve climate change all by themselves. Those are necessary, but not sufficient. We also need to figure out how to grow food without long-term desertification of the continent.
Interesting information! I really like the analysis of video content, particularly the "founder presence" info. However, I think that the headline results that isolate video cost when looking at raise results is a bit of an odd way to do this analysis because it leaves out the rest of the marketing mix. I'm actually preparing a crowdfunding campaign myself (first time), and I'd think that the main levers for success would be, in no order of importance:
A: Pre-campaign support (solid email list, social media presence, etc), B: Video Quality, C: Rest of Page quality (ie: copy, visuals, etc), D: Paid advertising, E: Media Coverage, F: (Most important) Compelling product & pricepoint
Money Raised should be a function of a-f (and several other factors I'm sure I'm missing) so isolating any one of those variables on its own is not that useful. Presumably a campaign with a $100k budget for its video has also made significant investments elsewhere.
Something I'm still trying to figure out: assuming there is a limited budget, would one rather spend 20k on a video or 5k on the video + 15k more on Facebook ads? My thinking is that the 3M extra views (assuming $5 cpm) would be worth more than the marginal increase in video quality from $5k-$20k.
Granted, if there's a good ROI on online ads, the budget there should theoretically be "unlimited", but that's not always the case. Curious about other peoples thoughts here.
Totally agree. I think that's what we are seeing from the other comments as well. Video is only part of the battle, and a more complete study would look at all the variables together.
Video is interesting in that it accelerates your efforts in other areas. The time you spend making a better video should also help you reach a bigger audience (A), increase conversion thus making advertising (D) more effective, it probably helps with PR too (E) since media outlets prefer to show interesting videos, and with page quality (C) to a degree since you can reuse frames from the video for the page.
So back to your question! how much to spend on video is you have $20k budget overall. From an execution standpoint, we certainly see a sweetspot ~$3-6k for video, in that it allows you to get a clean professional video that looks legit. That's enough to hire a director with a small crew for a day and simple editing. If you spend less then you'll have to take on some of the director's responsibilities to make sure you hit on the requirements we found in the benchmark (location scouting, quality images, tight narration...) We wrote more about this here https://www.videopixie.com/how-much-does-a-kickstarter-video...
Maybe 30% of graduates of top 30 universities. There are lots of CS grads from lower ranked schools that are definitely not getting fought over by the likes of Google/FB/Amazon...
Agreed. I've got an MSc, and the only contact I've ever had was a practically spam message for an Amazon recruiter (who contacted half the province I live in, as near as I can figure).
I'd be interested in seeing how a "single payer" approach to college financing would work.
High level -
The government pours many billions of dollars into colleges and universities every year, but because the money is disseminated via millions of individual loans to individual students, any buying power (and downward pressure on prices) that the government could provide is effectively eliminated. For a university this is a best case scenario: sell a "necessary" product to millions of individual customers with infinite money.
If instead, the universities could only "sell" to the government, the government would be able to actually demand meaningful concessions in price.
I'm thinking something like:
1. Students -$-> Gov't -$-> Universities (many more sellers than buyers)
Instead of:
2. Gov't -$-> Students -$-> Universities (many more buyers than sellers)
Easy to see how case 1 would keep prices lower in an ideal world...
It really feels like he's interacting with a person here. How can they do this??