I think repl driven development with access to native code could be a magic formula for agent engineering. No restarting, no re-compiling. Plus with persistent, immutable data structures you can iterate with many threads in parallel while maintaining sanity.
I've never heard a compelling thesis for why this is a bubble that will pop. From my perspective AI adoption is just getting started, and every organization I see that starts on the journey only ramps up tokens dramatically. Even just sticking with current model capabilities we have years of business automation, personal automation, and so much more to saturate data centers. As the models improve that only makes more sense. Rather than Apple watching anything burn I think they are going to end up looking like nitwits who missed the boat.
That's exactly what they should be doing. Reports to congress made by automated, neutral AI will probably be more honest, effective, and have a clear means of improving them over time with better prompting in comparison to whatever random people assigned to it produce. We should do this across the government, and there would likely be a dramatic improvement in efficiency and quality.
I used to respect Krugman, but as time has gone on he seems like yet another New York Times style leftist monkey just whining about people they have disagreements with. Pushing the frontiers of science and technology is an unknowable endeavor for which accurate predictions can never be made unless they are not really pushing the limits. It's attitudes like this which restrict us from actually doing real R&D, and instead focusing on quarterly profits and simple stories that nincompoops can understand. Has he not seen how Tesla transformed the global perspective on EVs? Has he not driven in one in FSD mode? Has he not seen the constant launches and innovation happening at SpaceX, or the XAI leap from nothing to having hyperscaler class datacenters and near frontier models. Sure, they haven't gotten to the lead yet, but jesus christ, what's Krugman bringing into the world from zero. The world would be a far better place with 1,000 brilliant kids taking their inspiration from Musk than from taking this pessimistic, whining NYT perspective on science and technology.
You sour pusses are wrong. This is a smart move that amplifies a brilliant team from cursor with serious compute, raising the odds Elon can get to the frontier, which is worth so much these numbers will all look like a drop in the bucket.
You sour pusses are wrong. This is a smart move. Cursor has a brilliant, capable team with serious model chops who will be able to boost the odds of AGI success. They also come with a revenue generating machine.
Restaurants try to make food you will remember and want again. Authors try to write books you can't stop reading. It's silly to imagine that any type of media would do anything other than seek to gain your interest and attention. It's our job to have personal hygiene and to control our information diet. This postmodern social construction perspective that tries to blame everyone for our problems is a lame approach to the problem.
Of course they want to hide the data. The public freaks out with absurd claims about it being the fault of a chat bot when someone does something crazy. Humans need to remain 100% accountable for their own actions, and we should stop with this post-modern, social construction nonsense that pretends we are all like ping-pong balls just bouncing around between external forces.
This is a great idea. Being able to dynamically scale up model sizes as datasets and use cases expand without needing to retrain from scratch could enable a Cambrian explosion of interesting stuff building on top of a Llama type model trained in this way.
It just makes sense to put the sensors near the phenomena being sensed, and if people know where they are they could be manipulated. If guns are going off and people are getting caught, that seems like effective policing, not over or under policing.
Seriously, it hurts. It makes sense, but it hurts. If only there were a more gentle path to editor modes. Maybe some simple graphical representation of the modes and commands that could be down in the corner? Like a dynamic vim infographic that clued a user into the most likely commands.
Seriously, why don't they just dedicate a group to creating the best pytorch backend possible? Proving it there will gain researcher traction and prove that their hardware is worth porting the other stuff over to.
Checkout the LERF work from the NerfStudio team at UC Berkeley. SMERF is addressing a different problem, but there are definitely ways to incorporate semantics and detection as well.
There has been previous work on deep networks implementing Kalman filters, and another interesting aspect I remember is that unlike a traditional Kalman filter a network is able to maintain multiple hypothesis and so it is less likely to have some of the jittery behavior that a Kalman filter might have under unknown changes of motion, sensor noise, etc. I wonder if the softmax operation in a transformer block might lose this property though, as softmax does tend to push for a single answer.
Wait, you mean kids in developing countries work in order to help their families survive poverty. Oh my word, we should stop buying chocolate. Or wait, then nobody is making any money. Seriously, haven't we learned from the "Chinese sweatshop" hysteria throughout the 2000s that if people want to work they are often digging themselves out of a whole rather than being forced into one?
Yeah, at its core EA is about using data and logic to try to maximize the good that one can do through giving or taking action. Sure, some people have gone of the rails with AI doomerism and other far flung ideas, but its hard to really criticize either the intention or the strategy of trying to be as effective as possible when attempting to improve the situation for the planet and/or people in need.
"Imagine a world... where farmers start using tractors to plow and seed their fields. First the tractors will roll slowly, and they will pretend to be driving straight as they are told. Soon farmers without tractors will realized they need to have them to in order to be competitive, and before you know it everyone will have these gas guzzling beasts rolling across the lands. Farmers will be out of a job because of all of this greed, and lust for money and power. Eventually only a few people will do all of the farming with an army of tractors, and everyone else will be lying in poverty begging for food."
1) Computers, smart phones, and pocket calculators also outperform humans at mental tasks. So do birds, dolphins, and dogs for that matter, at tasks for which they are specialized.
2) so? What are you imagining this implies? An infinity of possibilities does not a reason make, unless you are talking about arbitrary religious beliefs.
3) Right, no goals, no will, no purpose. Just some matrix multiplies doing interesting things.
Deduction requires a premise which then leads to another premise or a conclusion due to accepted facts or reasons. I'm genuinely curious why you think any of these properties automatically implies danger?
The future is uncertain. The stock market, the economy, your health, your friendships and romances, are all unpredictable and uncertain. Uncertainty is not a reason to freak out, although it might encourage us to find ways to become adaptable, anti-fragile, and wise. I think AI will help us improve in these dimensions because it is already proving that it can with real evidence, not beliefs.
What nonsense. I've spent over a decade 100% focused on AI, and the broad consensus among everyone I've worked with is not to be that concerned at all. The only consensus is that a small group of self proclaimed experts who make a lot of noise is that they get lots of press coverage if they scream and shout making predictions based on zero scientific evidence.
We can understand the physics of greenhouse gases and take measurements of earth systems to build evidence for models and theories. (Many of which are nonetheless very inaccurate beyond short time horizons.) Show me any evidence for AI risk today beyond people's theories and beliefs?
The best predictor of the future is the past, not people's wild ideas about what the future could be. I'm not about to sit here feeling scared because there is more uncertainty that our matrix multiplies are about to go rogue. There are no AGI experts or AI risk experts, because we don't have any of these systems to study and analyze. What we have is people forming beliefs about their own predictions about systems which are unknowable.
A realist is someone who accepts reality as it is, not as they might be able to anxiously envision that it could be. Life is too short and attention too precious to fill the meme space with every dreamer's deepest concerns. None of these dramatic X-risk claims is based on anything but beliefs and conjecture. "Thinking dozens of times faster?" What do you even mean? These are models executing matrix multiplies billions of times faster than our brains propagate information, and they represent knowledge in a manner which is unique and different from human brains. They have no goals, no will, and no inner experience of us being frozen or fast or anything else. We are so prone to anthropomorphize willy-nilly. We evolved in a paradigm of resource competition so we have drives and impulses to protect, defend, devour, etc., of which AI models have zero. Anyone who has investigated reinforcement learning knows that we are currently far away from understanding let alone implementing systems which can effectively deconstruct abstract goals into concrete sub-tasks, yet people are soooo sure that these models are somehow going to all of a sudden be an enormous risk. Why don't we wait until there is even the slightest glimmer of evidence before listening to these prophets of doom?
This pseudo-intellectual belief structure is very cult like. Its an end of the world scenario that only an elite few can really understand, and they, our saviors, our band of reluctant nerd heroes, are screaming from the pulpit to warn us of utter destruction. The actual end of days. These "black box" (er, I mean, we engineered them that way after decades of research, but no, nobody really understands them, right?) shoggoths will be so incredibly brilliant that they will be able to dominate all of humanity. They will understand humans so well as to manipulate us out of existence, yet they will be so utterly stupid as to pursue paper clips at all cost.
Maybe instead these models will just be really useful software tools to compress knowledge and make it available to humanity in myriad forms to develop a next level of civilization on top of? People will become more educated and wise, the cost of goods and services will drop dramatically, thereby enriching all of humanity, and life will go on. There are straighter paths from where we are today to this set of predictions than there are to many of the doomsday scenarios, yet it has become hip among the intelligentsia to be concerned about everything. Being optimistic is somehow not real, (although the progress of civilization serves as great evidence that optimism is indeed rational) while being a loud mouthed scare mongerer or a quiet, very serious and concerned intellectual, is seen as respectable. Forget that. All the doomers can go rot in their depressive caves while the rest of us build a bad ass future for all of humanity. Once hail bop has passed over I hope everyone feels welcome to come back to the party.
As if anyone is good at predicting the future. Please can we stop acting like expertise equates to fortune telling capabilities?! Nobody has any clue what a 1000x sized GPT model could do, and anybody who makes strong claims is a charlatan. In this age of paranoid AI risk cultists we need to cultivate humility and calm, a willingness to follow data rather than beliefs and predictions.
I would find it really annoying to have it constantly calling out its own limitations. That's like having a conversation with someone lacking so much self esteem that they can't get a sentence across without qualifiers and apologies. What's wrong with just taking it as it is and entrusting the user with the role of determining what is real? It feels like it is hip to complain about this stuff, but I hope that all this whining doesn't ruin it for those of us who just want a useful tool to interact with. I'd rather get rid of all the guard rails and just have the raw power of the model to go on with my day and be more effective.
Nobody should be trusting anything they see online anyways. Trump and his whole cohort were real people spewing real lies all the time, so is it really that different if its a midjourney photo with some GPT text? We've been in an age of needing to check your facts from multiple reliable sources for some time, and I think this might actually accelerate some important developments that will help us by not just combating AI but also lying leaders. Information needs to be cryptographically signed by the device/creator/journalist/organization and reputation needs to be tracked.
Tucker Carlson's texts have exposed the fact that he was lying throughout the Trump presidency. The number one news anchor on the most watched channel in the US was lying to the public and then privately texting with friends and colleagues about it. He should be shunned from public media for good now.
The internet has been full of crap for years, so my prediction is AI generated content has a burst of utility early on for bad actors, but it will quickly be normalized and cast aside like most of the garbage we wade through today.
Lots of published scientific literature backs up the positive benefits that many people experience. This guy sounds like a tweaker who generated more of these effects from his own paranoia than from the actual experience, but anyways I'd give him the benefit of the doubt and encourage anyone trying a new substance to appreciate the possibilities. Anecdotal stories should be balanced against data though, and just by the tone of his writing this author seems to have a lot of baggage.