artifaxx

Interesting, where do you think LLMs perform the best?

Tracking already feeling pervasive suffers from the cognitive bias of all or nothing thinking. A phone can be turned off or apps disabled far more easily than a network of surveillance cameras. There are degrees of surveillance and who has access to the data. We can push back.

Didn't lose me, but point taken about gathering more support. How about: the costs of implementing a zero-crime world are far greater than the crime. Or attempting to trade freedom for safety will result in losing both.

Definitely, AI sentiment is positive among most people at the small startup I work at in the Seattle area. I do see the "AI fatigue" too, I bet the majority is from using AI as a repeated layoff rationalization. Personally AI is a tool, one of the more useful ones (e.g. Claude and Gemini thinking models make quite helpful code reviewers once given a checklist) The hype often overshadows these benefits.

Indeed, there are levels to the asymmetry though. Oil change might be ~5x cheaper vs the 20-50x claimed for Heroku...

I can share a similar approach I'm finding beneficial. I add "Be direct and brutally honest in your feedback. Identify assumptions and cognitive biases to correct for." (I also add a compendium of cognitive biases and examples to the knowledge I give the LLM.

I don't see their comment as trolling. Reducing the negative impact Meta already has is a more accurate description. Given their track record so far, you'd be a tiny minority to even get to work on that instead of ad/user engagement optimization for example.

DDG is growing exponentially, Google is just massive and it takes a while to grow big enough to notice https://duckduckgo.com/traffic

Correlation vs causation leaves a lot open, chronically sick with a number of illnesses people may sleep more. Their mortality risk is also higher. I'd wish we could see controlled studies that randomly assign people to different sleep durations. You also have to control for gene mutations that affect sleep requirements https://www.scientificamerican.com/article/genetic-mutation-...

Different people are trying to get different things from Facebook. I use facebook but disable notifications and unliked all pages/companies. Crippling the annoying parts to them allows people to have their Facebook cake and eat it too. Without being bothered by stupid notifications mid bite.

I remember a professor who equated smartphone use with smoking, he told people to go outside with the smokers if they wanted to use their smartphones/Facebook.

This definitely depends on a lot of factors. If you do work where you are easily replaceable sure. Most software developers don't have anything special rights wise as so few unionize. As an independent developer if you choose work in niches where the demand is greater than supply you can fire your painful customers anytime and move on to other ones. Different solutions work for different people.

It really depends on the person, thats great you recognize that you do well in the corporate world. Unfortunately in many places the company would own your spare projects due to the intellectual property assignment rules in their employment contracts. Beyond that some of us strongly dislike conformity and being a replaceable cog in the corporate machine. Connecting with one's customers and seeing directly how what you build helps them instead of dealing with corporate politics is a huge improvement to me.

Depends on where you are, in some US states you can get licensed to call yourself a software engineer. This holds you legally liable to the code of ethics. It often influences court cases involving companies too: http://ethics.acm.org/code-of-ethics/using-the-code/

Software has a professional code of ethics, it isn't universally read/adhered to though: http://www.acm.org/about/se-code

Good point. I don't think Google did exactly that, but I did find this: https://hbr.org/2013/12/how-google-sold-its-engineers-on-man... It still is Google specific data though, so sadly it won't generalize as more than a set of ideas to test.

I agree, the title is certainly overconfident if not negligent. Preventing people from writing inaccurate communications would be the topic of many studies and papers by itself though!

For current issues yes, it doesn't matter you just fix it and get the results. If you want to prevent it in the future it helps to know which one causes the other. Then you can focus preventative efforts on the factor that causes the issue.

Unfortunately without putting randomly assigned teams into controlled settings to test this, how else can we figure that out? I want to know too, but this seems expensive to test to any high level of certainty for the causation.

I agree, its great to take that personal responsibility yourself. When it comes to software that others use, the easier you make it the more people use it. Isn't that a key part of why we use automation? Put work into setting it up once for end user simplicity and reap the rewards the next 10,1000,n times.

That is a good distinction to make, but if we don't know exactly what the cycle is things get harder to figure out. Neurons hold state and we aren't sure how much yet, let alone a deterministic method of computing it. https://www.scientificamerican.com/article/new-estimate-boos...

I agree the CAD analogy doesn't work well here. I don't think any of us can say for sure whether or not a high level representation will need to account for such details. You are right that it might not, but it sounds like it is a possibility worth considering

Wow, if mitochondria need to be modeled in that level of detail for accurate simulations of neurons that could be problematic. Exaflops of computational power were already the estimate just for simulating the neurons/synapses of a human brain[0], and this would increase that computational requirement orders of magnitude I imagine. Still quite fascinating. [0]http://www.riken.jp/en/pr/press/2013/20130802_1/

It can be healthy if done right http://www.mayoclinic.org/healthy-lifestyle/nutrition-and-he... It is probably a huge improvement over the crap you find in many hospital cafeterias.

It feels that way, but estimates look quite promising. If lab grown meat ends up using 1% of the land needed for animal production that would make a huge difference! http://phys.org/news/2011-06-lab-grown-meat-emissions-energy...

Testify makes a moderate improvement to writing tests, but I still love it. Multiply the improvements across a hundred tests and the difference is huge.

I find that yoga classes that integrate meditation (such as hatha) tend to work better. It also forces you to sit and work through it for at least an hour. Here is a page to get started with the evidence: http://www.health.harvard.edu/mind-and-mood/yoga-for-anxiety...

Where did you read this? I am sure I am not the only one here who would love to read more to get a better idea.

Thanks for a practical counterbalance to the pointless snark here. Your comments about software improving efficiency reminded me of Google's datacenter optimizations. https://www.google.com/about/datacenters/efficiency/internal... Data driven improvement can make a big difference in many places.

What many people don't realize is that you can put the paper in front of potential users and have them interact with it. You can draw/move the results of their actions to get an idea of how they will use your software before you have to write a single line of code. Getting inside the mind of your users and seeing their assumptions and desires in action is incredibly valuable.