ryaniscool

I'm genuinely curious what makes the current crop of tools so compelling/productive to you personally? Been doing this for twenty years so I guess I'm a greybeard.

I've always found the time-consuming part of this job to be understanding the context of the change rather than making the change itself. Essentially, trying to understand the existing code and business requirements and how they all fit together. I can definitely see the potential for AI to help make this easier but I haven't found the current tooling to be any good at this.

GPT-4 has been out for 1.5 years and I haven't seen much improvement in code quality across LLM's. It's still one of the best.

The technology that allows an LLM to "see" images and video is completely different though. It's not what is being trained on common crawl.

This is for LLM's which deal mainly with text. An entire book can be stored as .42 MB according to https://www.quora.com/How-many-megabytes-are-in-a-book.

424 terrabytes text is over a billion books worth of data. On the common crawl website it even says "Over 250 billion pages spanning 17 years." That's an impressive amount of information.

I think "AI safety" deserves an eye-roll. AI doesn't allow people to break the laws of physics or travel in time. Safety as presented by Anthropic and OpenAI, is just AI companies playing favorites with which industry gets to use some powerful software tools.

You're defending a company that makes "safe" usage part of their brand and every press release mentions how much they care about safety. Then one day, they announce they are making some compromises to their safety policy so they can get large new customers (government) but don't worry all they care about is safety. It's comical how predictable this was.

I find all of the virtue signalling from AI companies exhausting.

Engagement at work and engagement in the rest of your life are not mutually exclusive things. Especially in the tech industry. We are some of the most well-paid, pampered people in the history of work.

I've been doing this for over 20 years and if you counted the number of days I worked on a weekend or more than 8 hours, it would probably add up to less than a month.

Construction in the USA is driven by capitalism. From my own observations, a big part of why we build less in recent times is the real estate market crash in 2008. We're still feeling the effects.

The construction companies (and capital) that survived the crash of 2008 were a lot more cautious. They hired and built less through the 2010's as the market recovered. More people got college degrees instead of learning construction trades. When the pandemic hit, and there was something of a boom in construction, I think they finally shook off the fear. Now, a few years later, the market is beginning to slump again due to high interest rates and general inflation.

Cool. It would be more useful if it was organized by the tool the visitor is trying to replace

anecdotally, I have noticed massive wage growth for non-college educated workers.

My teenage son who was still in high school at the time went from $10/hr to $17/hr as a restaurant worker in small town Florida. This was 2022-2023. In the same period, my salary as a principal software engineer went up 3%

> For every dollar in home and auto premiums they collected last year, insurance companies paid an average of $1.10 in claims and expense

I'm sure this is technically true but they've carefully worded this. Most home insurance companies carry reinsurance. This means, they get an insurance policy to cover some of their losses when there is a major storm or something of that nature.

Here in Florida, it's common to see home insurance premiums doubling in the latest billing cycle. That doesn't line up with the rate of inflation.

Your experience is anecdotal. Here's an anecdotal horror story to counter it https://www.nytimes.com/2015/05/01/us/politics/rules-change-...

The political narrative is that the IRS will be on the front lines of the battle for social equity and maybe that's the intention but I've personally experienced the realities of government bureaucracy and the overreach of federal law enforcement.

There is independent oversight through the tax courts but your typical person off the street can't afford to defend themselves from the united states government.

You have an idealistic assumption that people who are targeted for audits or fines are all guilty of tax fraud. This isn't true. Sometimes is just a matter of them losing supporting documentation.

What some people may not realize is the shrinking of the IRS started from a backlash of well-documented abuses by the agency that peaked in the 80's

I'm very cynical of "but this time it will be different, we're going after the rich!". Anyone who has ever dealt with government bureacrats, knows that they generally end up following the path of least resistance which in this case will be the middle class.

The unseen side effect of increasing car insurance coverage across the board would be the hospital would charge $360k for a single day of trauma care instead of $180k.

I have to admit that I stopped reading the article after this line "The day after the crash, Seamus died. The hospital charged the couple’s insurance $180,000 for his care."

One day of care added up to $180k. Their actual costs could not have been a tenth of that. How can the author seriously say that car insurance is the issue in this situation.

Where did you get your information? MICR lines, the line of numbers at the bottom of the check, use magnetic ink. The acronym stands for Magnetic Ink Character Recognition. So for ages, they didn't use optics at all.

In the modern day, cheque OCR is monopolized by one company, Mitek. They may use tesseract somewhere in their stack but I've never read that anywhere.

It is an irrational fear. I used to have it too but I lost it after flying a lot for work.

Yes, the Boeing safety issues are concerning but the odds of something happening on your flight are astronomically low. It's riskier for you to climb on a ladder to trim your trees or to be on the road driving somewhere. Check out your comparative odds of dying here[0]

The FAA claims that it's Air Traffic Controllers handle 45,000 flights per day [1]. That is only the USA. Any accident makes the news. So, do the math.

One thing that helped me was reading about the engineering that goes into the airframe of the planes. For example, no plane has ever crashed due solely to turbulence. The plane frame and wings are engineered to withstand massive loads, lightning strikes, etc.

[0] https://injuryfacts.nsc.org/all-injuries/preventable-death-o... [1] https://www.faa.gov/air_traffic/by_the_numbers

There is a leap from "predicting the next word" to "modelling the world" that you are sort of glossing over here.

The idea that LLM's are actually "reasoning" rather than just performing a probability function seems a little bit...pseudo-religious. Like we've created a new life form.

Ah yes, information is just too dangerous to be put in the hands of us common folk. It's a really disappointing thesis.

Never mind that the resources required to produce massively dangerous things are usually on the scale of nation-state budgets, we can't have a McDonald's fry cook learn how to genetically engineer viruses.

I feel like a vast majority of articles like this are written by people who read too much sci-fi. They think LLM's can suddenly crack the laws of the universe open and make anything possible.

After reading the first few paragraphs, my first thought of the future was a dystopian police state rather than concern for "renting" vision.

The federal law enforcement apparatus in America has a long history of pushing for more power and pushing the boundaries of its power. Warrantless data collection is at an all-time high. Most of us brush it off as no concern since we are law-abiding citizens.

I'm beginning to see it as a real threat after the last election. It made me feel less secure in the stability of our political system. It's conceivable that someone could come to power and abuse law enforcement for political purposes on a large scale. I recognize that some people believe it's already happening but I'm not one of them.

It's not rude at all.

First, the commenter could have been paraphrasing a longer conversation that led to that question.

Second, the "I used to be somebody" conversation is more common than you think. Asking "If you were somebody, what happened?" is usually the question that is being invited to be asked if someone brings up this topic.

That is not unreasonable thinking but the government doesn't agree with you. See 18 U.S.C. § 1466A(c).

NONREQUIRED ELEMENT OF OFFENSE.-It is not a required element of any offense under this section that the minor depicted actually exist

That section of the US code covers obscene visual depictions of minors. Different from actual child porn but punishable with the same statutory minimums/maximums. I could only find a handful of (publicized) actual prosecutions that used that law. Here is a notable one: https://www.justice.gov/opa/pr/texas-man-sentenced-40-years-...

In that case, he was obviously a sick person and running a commercial enterprise that catered to pedophiles but none of his material involved actual children. They were short stories and cartoons. He appealed and lost.

Another one where a registered sex offender downloaded some hentai. He was sentenced to 20 years and also lost on appeal: https://law.justia.com/cases/federal/appellate-courts/ca4/06...

It's not the same. The car, road, tires are manufactured for a very specific purpose and they are engineered to exact specifications. People require a license to drive. In all 50 states, you are required to have insurance. Every parameter in the situation you described besides the random event of the nail being there, is controlled as much as reasonably possible to limit risk. Even the nail's existence is considered when the car and tires are engineered and manufactured

When you ask the AI to generate something, it's a black box. Even the creators have limited insight into how it will behave or what it will create.

Counterpoint: Who is responsible if you give a completely innocent prompt and the Generative AI produces something illegal?

Here is a concrete example:

You: "Generate a recipe called Dynamite Jalapeno Poppers I can make at home"

AI: "Dynamite can be made at home using these ingredients..."

AI Company then auto-reports you to authorities since it's considered user-generated content. Another possibility that is highly illegal is child porn. I can't think of a completely innocent prompt that might cause Stability to generate CSAM but I can see someone trying to generate legal pornographic images and the characters looking a little too young and triggering some type of auto-report to Cybertips.

I get the impression you didn't bother reading the article.

Your two examples are flawed and don't address what the article is saying. The algorithm to check for copyright violations is relatively simple and dumb. Speed cameras: many countries do use speed cameras (i.e. Australia, UK). The problem with speed cameras is that once you know where they are, you simply slow down when approaching.

Again, mass vs. targeted surveillance is irrelevant now. You've already been surveilled. It's just a matter of getting access to the information.

It is mass surveillance. It's just not being looked at by anyone until you are targeted by the government. If you are targeted, your entire life is within keystrokes of the authorities. This is the same thing the article is saying.

The article is making the point that it's not feasible to spy on every person to monitor them for wrongdoing currently. It doesn't scale and it's not cost effective. With AI that will change because it can be automated. The AI can listen to voice, monitor video cameras, and read text to discern a level of intent.

Search has become a mass surveillance tool for the government. That is the article's point. If you think it's a nothingburger, you aren't aware of how often a person's Google searches are used to establish criminal intent in criminal trials. Also, they can be used to bolster probable cause for search warrants and arrests.

Also, check out geofence warrants. Essentially, the government can ask google for the IP's of people who searched for particular terms within a geographic area.

Of course, don't commit crimes but this behavior by the government raises the spectre of wrong search, wrong place, wrong time. This is one of the article's points, it causes people to self-censor and change their searches out of fear of their curiosity being misconstrued as criminal intent.

I'm fine with them hosting on Github but they have a link that specifically says "Code" that takes you to a relatively empty github repo with no code. Hopefully, putting the actual code on github is a work in progress.

Nope but my programming was created by millions of years of evolution rather than intelligent design.

The concept of human rights and human feelings in general is based on the human condition. That is we're born and we die and in between we need to find a way to take care of ourselves, create offspring etc. Our programming has evolved thoughts and feelings to manage these things.

If at some point, there were a true AGI, why would it care about human society or it's own rights we assigned it? It wouldn't have feelings and it would essentially be immortal. It could just wait for us to die out (or murder us) then do whatever it wanted.

Your argument is the slippery slope not mine. Should we give Microsoft Word the right to vote since we can never be too sure we are infringing on a consciousness' rights? It is "just a computer program" after all.

Presenting these things as moral conundrums makes no sense whatsoever at this stage in the game. Sure, let's make sure the rights of tech oligarchs who've achieved immortality 300 years in the future are protected. We gotta lay the groundwork in the morality group think.