0 Members and 2 Guests are viewing this topic.
Why bother? The real universe provides all the facts and tests you need to do anything.
There have been 4 research papers and technological advancements over the last 4 weeks that in combination drastically changed my outlook on the AGI timeline.GPT-4 can teach itself to become better through self reflection, learn tools with minimal demonstrations, it can act as a central brain and outsource tasks to other models (HuggingGPT) and it can behave as an autonomous agent that can pursue a multi-step goal without human intervention (Auto-GPT). It is not an overstatement that there are already Sparks of AGI.
Requirements for criminal evidence will be different from now on.
Why should the requirements vary?
Can GPT 4 Prompt itself? Give it a mission and it will come up with the prompts. This video showcases the rise of autonomous AI, including 5 major developments in the last 48 hours. Starting with the OG Auto-GPT, we see how it quickly gain text-to-speech, coding and more. Karpathy weighs in and then we see how you can now create an app with just your voice, with a Jarvis demo and another route via Imagica.AI. I then showcase MemoryGPT, a brand new model that can permanently store previous conversations and remembers topics the next time you ask.I also cover the concerningly rise of models such as ChaosGPT that show how people will create malicious goal-seeking models just for fun. The video shows how you can now create a shareable bot on poe.com, with any personality you like (images were from Midjourney v5) and what Anthropic are working on with Claude Next [plus Nvidia million x quote].You'll see how Microsoft Jarvis, using HuggingGPT, is hit and miss, and how Sebastien Bubeck shows we are not even seeing the raw potential of GPT 4. I end with a disagreement between Yudkowsky and Altman, via Baby AGI, on whether we can use AGI to align AGI.
Ahem Ahem!Soo then...What happens when A.G.I..Starts creating Porn?(not trying to derail or hijack the OP, it's a genuine serious question)
AI researchers are worrying about alignment problem. IMO, it can be broken down into several parts:- Identifying the Universal Terminal Goal.
- Describing objective reality as accurate and precise as possible, including causality and relationships among objects.
- Identifying necessary actions to achieve the universal terminal goal based on best model of reality that we currently have, which include setting general rules and instrumental goals.
There isn't one, unless you are a bee or an ant.
What we now have is a machine that finds apparent correlations between published documents. Nothing more.
What's the universal terminal goal, if I were a bee or an ant?
Quote from: alancalverd on 13/04/2023 09:23:39What we now have is a machine that finds apparent correlations between published documents. Nothing more.Some machines have already analyzing CCTV records, as well as financial transactions, which carry important facts about human society. There are also machines to analyze comments on social media and browsing history of internet users.
Exactly my point - it's all old data in the public domain.
Quote from: hamdani yusuf on 13/04/2023 15:13:25What's the universal terminal goal, if I were a bee or an ant?Survival of the nest. Occasionally there will be something of an upheaval triggered by the queen's hormones but unlike more stupid species it doesn't lead to civil war, just a relocation and a new queen, with the same goal.
https://www.quora.com/Is-it-true-that-some-worker-bees-are-able-to-lay-eggs-which-hatch-as-drones-if-the-queen-diesSometimes you may have a laying worker even if you have a queen. Queens tend to lay solid, tight brood patterns. Laying workers are more haphazard, and yes, in most cases, they will only produce drones. Researchers estimate that a little less than 1% of workers can lay female eggs. Since drones contribute nothing to the health and welfare of the hive, they are a drain on the hives resources. That is why the worker bees will often kick out the drones when Fall comes—when times are tough, you don’t get to eat if you don’t work. Since drones don’t work, a hive with a laying worker will eventually die off. And if you’re into scaring yourself with bees, read up about the Cape honeybee. Apis mellifera capensis Escholtz — female workers can lay eggs that develop into drones, workers, or even queens. They are a difficult to manage race of bees and will parasitize other hives and take them over, fight among themselves in line with family loyalties (daughters of one mother will fight daughters of that mother’s sister, and both groups may reside in the same hive at the same time.) The problem is that they don’t need a queen in order to take over . . . a laying worker will do just fine. Big challenge.
It's still secondhand data. The idea of successive approximation is not new. I have two flight directors on my plane - an old analog "localiser" that gets more sensitive as you approach the target because it measures angular deviation from a radio beam, and a modern GPS digital system that expands the display scale every mile. No hint of intelligence or creativity.
What new is the capabilities of machine to understand context and deeper meaning of inputs with various modalities, and make conclusions from those. They are not perfect yet, just like little human kids. But they can improve at much higher speed than any human can.