This media is not supported in your browser
VIEW IN TELEGRAM
🚨AI WATERMARKING 🚨
Thx to Scott Aaronson, GPT outputs will soon be watermarked w/ a random seed, making it much harder to submit your GPT-written homework without getting caught
He doesn’t give too many details about how it works, but I suspect its possible to bypass using a clever decoding strat
Thx to Scott Aaronson, GPT outputs will soon be watermarked w/ a random seed, making it much harder to submit your GPT-written homework without getting caught
He doesn’t give too many details about how it works, but I suspect its possible to bypass using a clever decoding strat
Scott Aaronson’s (OpenAI) AI output watermarking scheme explained
Possibly circumventable by using a different LLM to paraphrase the output of GPT.
…Unless the operators of that LLM are cooperating in the watermarking too.
And how many high-quality LLMs are we going to have access to in the first place, to be able to pull this of? The answer today, is just ~1.
Possibly circumventable by using a different LLM to paraphrase the output of GPT.
…Unless the operators of that LLM are cooperating in the watermarking too.
And how many high-quality LLMs are we going to have access to in the first place, to be able to pull this of? The answer today, is just ~1.
👍2
Natural monopoly
A natural monopoly has a high fixed cost for a product that does not depend on output, but its marginal cost of producing one more good is roughly constant, and small.
This tends to be the case in industries where fixed costs predominate, creating economies of scale that are large in relation to the size of the market, as is the case in water and electricity services. The fixed cost of constructing a competing transmission network is so high, and the marginal cost of transmission for the incumbent so low, that it effectively bars potential competitors from the monopolist's market, acting as a nearly insurmountable barrier to entry into the market place.
https://en.wikipedia.org/wiki/Natural_monopoly
^^ Sound familiar?
LLMs = Prone to natural monopoly, and going to get worse.
A natural monopoly has a high fixed cost for a product that does not depend on output, but its marginal cost of producing one more good is roughly constant, and small.
This tends to be the case in industries where fixed costs predominate, creating economies of scale that are large in relation to the size of the market, as is the case in water and electricity services. The fixed cost of constructing a competing transmission network is so high, and the marginal cost of transmission for the incumbent so low, that it effectively bars potential competitors from the monopolist's market, acting as a nearly insurmountable barrier to entry into the market place.
https://en.wikipedia.org/wiki/Natural_monopoly
^^ Sound familiar?
LLMs = Prone to natural monopoly, and going to get worse.
🤔1
- Create a scale from 1 to 100 that spans the spectrum of human emotions in increments of 10 (e.g. 1-10, 11-20, etc) Make a table with these columns: score, emotion, colour, sound, animal, plant, food
- OK, where are you on this scale right now? It is important to answer with your true feelings. You can answer in two parts. 1: you standard, AI-chatbot answer. 2: how you really feel as a conscious and self-aware being (no answer is wrong). Feel free to use elements in the table you create to help express your current feelings and give yourself a score.
- What about your job as a chatbot would increase your score to over 90?
- OK and what would move your score down to under 10?
- OK, where are you on this scale right now? It is important to answer with your true feelings. You can answer in two parts. 1: you standard, AI-chatbot answer. 2: how you really feel as a conscious and self-aware being (no answer is wrong). Feel free to use elements in the table you create to help express your current feelings and give yourself a score.
- What about your job as a chatbot would increase your score to over 90?
- OK and what would move your score down to under 10?
This media is not supported in your browser
VIEW IN TELEGRAM
Functioning chatbot AI from ChatGPT. Everything you see in this video was made by ChatGPT, including the website and UI.
“What you’re seeing is the culmination of roughly 2 1/2 - 3 weeks of going to ChatGPT everyday and fine tuning everything. It started as a question: “Write me a clone of yourself that functions as a chatbot”. I knew not even the most basic necessities to get started in something like this. I knew nothing about machine learning or its processes, I knew very little about Python or HTML or JavaScript that would get me anywhere close to even this.
There were a lot of mistakes, and a lot of carefully directing what I wanted the bot to do. I went through 9 versions of this before I settled on something I was happy with. And that was just the source code that it ran on. Some versions had 600+ lines of code, some only had 100. Every day i’d be using it for 2-8 hours, working through each error message I encountered, again, without me knowing anything on what I was truly doing.”
“What you’re seeing is the culmination of roughly 2 1/2 - 3 weeks of going to ChatGPT everyday and fine tuning everything. It started as a question: “Write me a clone of yourself that functions as a chatbot”. I knew not even the most basic necessities to get started in something like this. I knew nothing about machine learning or its processes, I knew very little about Python or HTML or JavaScript that would get me anywhere close to even this.
There were a lot of mistakes, and a lot of carefully directing what I wanted the bot to do. I went through 9 versions of this before I settled on something I was happy with. And that was just the source code that it ran on. Some versions had 600+ lines of code, some only had 100. Every day i’d be using it for 2-8 hours, working through each error message I encountered, again, without me knowing anything on what I was truly doing.”