With a positive expected value, in theory, so long as we’re able to sustain growth forever.

I think you’re just asking when people form their first coherent memories.

This is about 2 to 3 years old.

Thinner until I can shave with this thing.

They don’t need to run failure studies. That basic characterization work is mostly already done for you. Component vendors (should) publish tables of mean time between failure for the components you’re buying that can be used to get a rough estimation with just a few minutes of effort. Typically it’s indexed by temperature, like for caps, but depends on the part.

Now, does the bottom of the market actually use those tables? I can’t say for sure. I know one high power headlamp company does this for their LED drivers to balance lifetime with output, but I can’t know for sure what every business does.

The LED bulb itself is not typically the reason that these fail. There’s also complex regulation circuitry which consists of far more complex components than a single LED. Electrolytic capacitors and driver circuits can have shorter lifetimes than the actual bulb.

With that said, manufacturers know this, so they also tend to overdrive lower cost LEDs to bring the failure rates in line with the rest of the circuit. This sounds like this may be what has happened to you just based on the dimming, but without knowing exactly how they have wired it up, it’s difficult to be sure.

And that’s fantastic! That’s what technology is supposed to do IMHO - Give you more free time because of that efficiency. That’s technology making life better for humans. I’m glad that you’re experiencing that.

If they’re not hallucinating as you use them, then I’m afraid we just have different experiences. Perhaps you’re using better models or you’re using your tools more effectively than I am. In that case, I must respect that you are having a different and equally legitimate experience.

I do have a concern for the health of the overall ecosystem though. Don’t all good devs start out as bad ones? There still needs to be a reasonable on-ramp for these people.

It’s rare to see such a complete and well-thought-out response anywhere on the Internet. Great job in capturing the nuance. It’s a powerful and often-misused tool.

It sounds to me like you’ve got a good head on your shoulders and you’re actually using the tool effectively. You’re keeping yourself in control and using it to expand your own capabilities, not offloading your job responsibilities, which is how more inept management views AI.

I think your question is covered by the original commentator. They do hallucinate often, and the job does become using the tool more effectively which includes capturing and correcting those errors.

Naturally, greater efficiency is an element of job reduction. They can be both hallucinating often and creating additional efficiency that reduces jobs.

Still a bargain.

Far be it from me to defend the Nazi car company, however:

While the results may seem like a damning indictment of Tesla, the report notes that the company has improved the build quality of its vehicles. All of its latest models now offer “better-than-average reliability,” and Tesla ranks among the top 10 brands in Consumer Reports’ new car predictability rankings, surpassing established automakers like Ford, Chevrolet, Mercedes-Benz, Audi, and Volkswagen.

A significant contribution to that is residential power effectively subsidizes data centers for AI.

Felons doing felon things. Have we no self-respect?

I’ll bet you $100 what side Congress will take in the matter, and it will be on the side of money.

I look forward to my first $400 power bill.

You don’t get to order people who don’t work for you.

We’ve got representatives attacking the postal service for not being highly profitable. We aren’t at square one of investing in the public good even when it pays in the long run.

Costs are borne by all, but profits only for the few.

My dumbass wondering what a Tacophone is.

Then, he is a fool. LLM technology has no fence around it. You can download and run one on your own hardware. The only reason a person would use their service is convenience access to a larger model.