jameslevy

Is that facebook thread considered terrifying because first car is often a security question? Or something else?

The only longterm solution that makes sense is to allow students to use AI tools and to require a log provided by the AI tool to be provided. Adjust the assignment accordingly and use custom system prompts for the AI tools so that the students are both learning about the underlying subject and also learning how to effectively use AI tools.

Does it require Azure to work? Or could the Azure steps be relatively easily be swapped out for AWS/GCP/etc?

Squiggle is an extension of the OpenAI Realtime API that generates insights (summarization, highlights, structured data, and content flags) in realtime as the conversation occurs. You can use Squiggle with a one line code change, no account setup needed.

The website has a link to a simple demo and an embedded video of the demo in action.

I originally built this for something I'm working on, but figured I should make it available as an independent service that anyone can use.

The insight-generation features do incur additional costs utilizing the OpenAI API, with the same OpenAI API key you are using for the Realtime API, and you can configure how often the insights are generated and thus how many extra API calls are being made.

Any feedback is very appreciated!

First money in at a 40M valuation? Assuming this is pre-seed, that seems like the investor is taking on a lot of risk based on the distribution of outcomes.

"Show me the incentive, I'll show you the outcome". I do not know anything of the culture of Tesla and am not commenting on it specifically, and that quote could apply just as much to a company like Boeing. But at companies working on safety-critical products, the problems usually arise from employees acting in their own self interest to do what will get them rewarded for shipping on time, or to avoid punishment for causing delays.

Goodhart's law applies to how inflation measures have ceased to be an accurate measure: "when a measure becomes a target, it ceases to be a good measure."

It's surprising how many parents I know who are in denial about this. It must be because they themselves are constantly using social media and don't want to accept what it is doing to their own mental health.

These types of tools should exclude all files and directories in the .gitignore as standard operating procedure, unless those files are specifically included. Not just because of secrets, but also because these files are not considered to be part of the repository source code and it would be unusual to need to access them for most tasks.

It's amazing to me how often Goodhart's Law can be observed in the wild, and I've come to believe it is an important enough concept that it should probably be something (along with logical fallacies and other concepts related to critical thinking) that should be prioritized above a lot of other things taught in high schools.

What about the MMT concept of just taxing the rich and "burning" their funds as a targeted method of reducing the money supply?

Perhaps prompting the model to first describe its approach to answering the question. This type of chain-of-thought technique can yield better results.

Hopefully they will publicly commit to only using this trademark defensively if granted. In other words, to give OpenAI the benefit of the doubt here, they may just want to prevent someone else from trademarking "GPT" and taking legal action against them.