MindFactory @MindFactoryAI

@ricwe123 @sama Did it really though?

@chris_j_paxton It doesn't really matter to be honest. AI is just not that impactful yet. We are still in the 286 phase with the 386 only on the horizon.

@jhall @ylecun @elonmusk Fiery but mostly peaceful. Biden is the sharpest he's ever been. The laptop was disinformation. You can't spread or catch COVID if you are vaccinated... etc etc The list of outright lies committed by liberal, supposedly educated journalists grows every day.

@ylecun @elonmusk I guess you haven't been in the US long enough to understand how the media works here!

@SaindoMatrix @showerskittles @ylecun @elonmusk Nobody cares about Jan 6th bro. They care about inflation. The problem with people like you is you mistake what you read on social media and see on TV for things that people actually care about.

@showerskittles @ylecun @elonmusk Well said. There is a big difference between "teh science" proposed by people like LeCunn, which is nothing more than an appeal to authority, and actual science.. which is adhering to a philosophy of providing disprovable hypothesis and evidence. LeCunn is cheapening science.

@CyberPunkCortes @pwrflwzrd It seems that there would need to be a set of escalatory steps, before total war between nuclear powers could occur. In fact, the more I study these wars, the more I come to the conclusion that states are rational actors, despite propaganda stating otherwise.

@CyberPunkCortes @pwrflwzrd Nukes are bad, sure. However I think after all this time it’s pretty clear that MAD works, thus the lack of concern. In fact, Ukraine reveals that the massive collateral damage nukes cause is a liability, not an asset in modern warfare.

@bryancsk High migration rates, underinvestment in infrastructure and a culture of real estate as a financial investment instead of a place to live. Space is not a problem, there is a lot of cheap land. However nobody wants to live where there is no mains water, sewage, roads, etc

@tillathenun @GadgetsBoy Try being a workman who is paid by the hour. That might make you “impatient”. EV is a luxury good for most workers. EV are fine for short trips between recharge stations, that’s it. Writing off valid concerns about resource usage because they are new is deflecting.

@DrJimFan @vkhosla It's a self referential definition. A job is a contract that occurs between two humans. A job, by definition will always be something done by a human.

@Ilir_AI More dexterous that digit? Wrong!

Sensor upgrade for the aloha system. Oak-D Pro RGBD (structured light)

We are organizing @corl_conf 2023 workshop on Reliable and Deployable Learning-Based Robotic Systems with an exciting list of invited speakers, looking towards the future of robot learning systems: corl2023deployable.github.io. Please don't hesitate to submit your work here!

Making a slight change to creator monetization: Any posts that are corrected by @CommunityNotes become ineligible for revenue share. The idea is to maximize the incentive for accuracy over sensationalism.

Today my garage research is to quantize a small local LLM ChatGPT-like model called: CognitiveBiasCheckerGPT It runs on a Raspberry PI and you can send any text from news or other sources and get marked up Newspaper-like output showing the 150 cognitive biases. I will open source this model and have a ready-made flash for the SD Card. So why is it on a separate device? There are many reasons. But one is that you can direct any text or have it fetch text from the web (soon any audio) and be anywhere in the world. It will rate each article and keep a database if you like of the likely biases it detects. You will have an export feature so you can share and merge this with other users. You will never read anything the same way again. Private and personal local AI is your super power in an over saturated world of “look here, not there”. I will update as I make progress in my garage labs.

Finding grasps that enable robots to interact with articulated objects is challenging because: 🗄️articulated objects exist in infinite joint configurations 🤖grasps need to be both stable and actionable We tackle this in our work AO-Grasp 🔗 stanford-iprl-lab.github.io/ao-grasp/ [1/7] 🧵

@tonyzzhao @lachygroom Also, you could check out my review of the paper on YouTube. youtube.com/watch?v=N5xWeh…

@linylinx Sounds like the CEO just didn't know basic security practices. Being any kind of foreign national is going to be a red flag if the company requires you to get security clearances. As an Australian working in USA, I wouldn't bother applying to defense contractors.

@Scobleizer @DrJimFan @tegmark The above could be a world model, but you would have to do F([state_token, state_token.... action_token] = [state_token, state_token, state_token] However the dataset really isn't structed that way, so you wont get good predictions.

@Scobleizer @DrJimFan @tegmark That's not a world model. It's a cool prediction, but not a world model. World model F(s, a) = s_next where given a state of the world, and an action, I predict the next state. Above you have F([token, token ... ]) = next_token