ChatGPT Opens A Research Lab…For $2!

Here is a crazy idea. Let's slice up a brain into small pieces. Okay, let's rewind a little. First, let's not just use ChatGPT, but use it to create a full research lab. Now wait a second. That is of course impossible. A research lab requires the work of several people. How would we do that with just ChatGPT? Well, here is an even crazier idea. Let's create several copies of ChatGPT and ask them to pretend to be a professor, a PhD student, a software engineer and more. Is that possible? Well, have a look at this. Previously it worked with video game characters. They made 25 ChatGPT agents, gave them motivations and memory and put them in a simulated town. They wake up, start their morning routine. Some of them like reading papers. I approve that. And they held elections too. Wow. What happened there? Well, Tom, one of the characters said, to be honest, I don't like Sam Moore. I think he's out of touch with the community and doesn't have our best interests at heart. End quote. Oh my. I wish the average voter sounded like that too. So I show you this because they can have relationships with each other and even help each other. So in this new work, they made a research lab to work on tough research questions. And it did not go as I expected it. Not at all. Dear fellow scholars, this is two minute papers with Dr. Caro Zsolnai. So it has its own research staff? Yes, but that is not where we start. Look, we start with a human putting in an idea. Very important detail. It starts with you. The idea, for instance, can do biases affect the performance of language models on benchmarks. Now a PhD student, which is played by ChatGPT, is asked to look up if the idea has already been solved before. Then a more experienced postdoc Researcher, another. Another ChatGPT, is involved with making a plan. And finally, two more little AIs are asked to code the whole thing up. Note that there are plenty of moving parts here to make sure we get a quality work. So then what happened? Well, absolute magic happened. The concept wins medals left and right. It outperforms previous techniques on a variety of tasks. Just make sure not to ask about anything that is in Russian for some reason. Now, about that brain. I think this is absolutely incredible. It is like taking a huge brain and instead of just using that, we slice it up into smaller little brains and that somehow does better. This is insanity. What a time to be alive. Yes, this is Two Minute Papers, the corner of the Internet where we slice brains up and we are Even happy about it. So how much does the slicing cost? How much do we need to pay our little AI researchers? And now hold on to your papers fellow scholars, because this new technique can do all this for the grand sum of $2.33 and it is done in 20 minutes. And if you wish to plug in these Fancy new thinking AIs, they do the literature review a bit better. I love seeing that nice gradient downwards here. This takes a bit longer and costs a bit more. But even if you wish to feel like a king and go all out, you pay about $13. And in this case it takes one and a half hours for proper research work with an implementation, experiments and more. Crazy. And whenever I don't have the resources to do that, I just rent a GPU on Lambda and do it myself. By the way, the full code and the paper is available for free for everyone. So this is open science lo and there is a more detailed user study in the paper. The link is in the description. And this fellow scholars, is how I see the future of AI. And that is empowering human thought. This is meant to help you with time intensive repetitive tasks. But ultimately you are in charge. Now one super important question. Can these AIs really invent something fundamentally new? Lets see. Well their ideas tend to be more novel than ideas of humans. That is good, more exciting, even super good. But here come the bad news for the AIs. Their ideas were evaluated to be less feasible at the same time. I mean it is easy to come up with super exciting things. Teleportation, time machines, anything you want. But ultimately the ideas have to be practical too. So I would give this one to humans. And that is why the idea has to start with you. For instance, the Nobel Prize winning Alphafold was not just about throwing an AI at protein folding by far. No, no, no. It required the ingenuity of tons of brilliant research scientists and the interplay of many, many moving parts. So the AI does not invent fundamentally new things unless it meets human brilliance. So what do you fellow scholars think? Let me comments below. Would you like to run your own copy of Deepseek in the cloud cheaply without using the official app? Yes. Then try Lambda GPU cloud. They have so many powerful Nvidia GPUs with tons of memory. I use them regularly too. Seriously, try it out now@lambdalabs.com papers or click the link in the description.