38

You don’t need to delete your posts… I’m not offended by ur constructive criticism. I was just saying ur way of writing was a bit off-putting.

Sorry I missed this. Yes, those topics are largely unpredictable with any level of accuracy. Furthermore, the world is not just an algorithm to be optimized.

I think you might have miscontrued what I was saying. Yes I agree weather, volcanic eruptions, large large large natural systems are “quite unpredictable”. In fact, vast arrays of studies show this to be the case. Chaos rules, and the butterfly effect is extremely problematic when small initial conditions change. HOWEVER, recent advances in fluid dynamics modelling will solve some of these issues. DMD (Dynamic Mode Decomposition see https://youtu.be/sQvrK8AGCAo) was used extensively for fluid dynamics, climate change modelling, mouse brain modelling and so on. The new ODE45 intertwined with MLPs seem fascinating (see https://youtu.be/jltgNGt8Lpg). Likewise Multiscale predictions (see https://youtu.be/Jfl3dIlSTrU)

I once again AGREE with you that large scale natural systems are very very very very very hard to predict.

In all cases, you had high quality data handed to you. The real world is messy.

Oh no nononooo. The data was MESSY AS HELL!!! We had to predict rainfall patterns in Pakistan in collaboration with UNSW researchers, since large downpours caused flash flooding causing death. The data was ATTROCIOUS! People accidentally put data in the WRONG columns, averaged the results up wrong, whole swathes of data from regions were wrong, data averaging across time didn’t match with the averages the people gave etc. It was a nightmare.

Predicting Zika Virus outbreaks in collab with the National Health Medical Research Council and EpiWatch was also very nightmarish. Data exists yes, for cases of Zika. But the issue was to predict the ONSET of an outbreak. So what data do we use? We used Google Trends in multiple languages for every state in Brazil. But it was still very weak. So we had to devise a new model entirely from scratch called the Markovian SIRD model, where we took into account transport links, airplane routes, road links, population, mosquito population distributions, temperature, humidity, etc.

That’s an arbitrary definition that you just invented. The definition of AGI is that it “can perform any intellectual task that any human can”. Can humans predict the future with extreme accuracy? No they cannot. You’ve arbitrarily set the goalpost for AGI to something that is literally impossible. Humans can, however, predict the future based on passed experience. We do this all the time when we predict what word someone is going to type ____. Did you guess “next”?

Ok I agree. I accidentally conflated AGI and ASI. Superintelligence might not even be able to answer future prediction questions, although I would really want it to. AGI fine, using a baseline definition, AGI should do everything a human can do intelluctually (let’s discount physical movements for now). I would even say an AGI should just do 51% better than the AVERAGE human, and I would be happy. Yes I agree a basic AGI system based on say the Mixture of Experts approach should be able to exist in the next few years. It seems reasonable.

Yes, it sounds like we agree. I just don’t understand why you’re wasting your time and talent on this problem then. Don’t you think that tech giants like IBM, Amazon, and Microsoft would be working on this problem with considerably more resources than you?

Because it’s what makes me not sleep at night!!! I LOVE what I do? Is that a good enough reason? I super extremely just super duper love what I do everyday, and I’m just happy I’m trying to make a good difference to the world. The financial predictions - I agree competition is rife. But so? My family is damn waiting for me to manage their funds. Likewise, life expectancy, disease, flood, fire risks, house prices, heart disease etc prevalance - these might exist ok yes. But once again - I’m waiting to use it. I want to know how long I live. I want to know how to improve optimally. I want to know what factors I need to look out for. My family has a long history of hypertension, stomach cancer and my grandparents siblings have alzhemiers. I want to know how to live longer. Don’t we all?

At the end of the day, I understand fully and I agree with you - “but isn’t OpenAI, Microsoft, Deepmind, governments from the US, China doing this?” Yes I had that mantra before. But it doesn’t faze me. Even if other large corporations will attempt it, I’m sure what Moonshot does will help the world. Even if we’re stamped and trampled on, I want to do what’s right for the world, and not regret any decision I make when I die. I believe that we’re aiming for the stars, and it’ll be tough - super tough. But I want to press on!!!

Anyways thanks though @daveshapautomator for ur precious time. I wholeheartedly thank you for actually making you write a lot, and I appreciate it. Obvisously some rephrasement might be necessary, but in general your criticism is just and reasonable.

Yikes I wrote too much.

39

Homegrown ai will out preform in other sectors of wellbeing, and so far all of them work hand in hand, the good ones at least, so keep up the homegrown work @danielhanchen , I’ll be routing your Jarvis some day,

1 Like
40

Thanks @YT !!! Super appreciate your support and thanks so much!! I really hope Moonshot will help make the world a better place!!

Yess!! When a basic JARVIS version comes out, I’ll post it here!! :slight_smile:

1 Like
41

Well, I completely understand intense passion. There were many days that I woke up at 4am to work on my AGI and write my book. And it wasn’t by choice either. Carry on.

FYI I was under the impression that you were trying to start a business, hence my pointed questions. That was the source of disconnect.

1 Like
42

Your questions are fine in terms of content. Possibly I over-interpreted / over-analysed the exact underpinnings of ur phrasing.

But anyways once again I’ll defs try reading your book! You’ll have to give me some time though since it’s been real hectic recently!

Ye in the business world, competition is rife. If I can’t take pointed questions, then I guess I deserve to be culled :slight_smile: Definitely you made me stronger and more resilient - so once again thanks @daveshapautomator !

1 Like
43

My brother made a delicious stunning website for Moonshot!!!
www.moonshotai.org

We’re planning to probably get a better domain, but for now it’ll do!

Any suggestions are welcome!! The site is still under HEAVY construction, so buttons won’t work, pages are blank / still being filled etc!

A small snapshot of the site!

image
image615×677 121 KB

3 Likes
44

With your expertise in algorithm optimization I was wondering if you had any thoughts about memory search? I am aware of tools such as ElasticSearch and Pinecone, which are great for rapidly searching vast amounts of text. I believe these technologies are “good enough” for now, but I feel like we will also need a way to compress, summarize, and extract relevant information from memories. Sure, we can do this with the Answers endpoint right now but it is inefficient, slow, and expensive to do so. It’s also not very good, IMHO.

Right now, my best bet is to use a fine-tuned model for handling memories. This model could be used to do various things such as answer questions about a given memory, summarize a memory, and so on. I’ve even imagined that we might need a “BS detector” fine-tuned model. This would emulate a neurological function that happens in our brains. For instance, if you hear something that clashes with what you believe to be true, you get a very distinctive brainwave patter - a rejection signal. The need for this arises because without it, an AGI entity might take what everyone says as true and factual, even if it is confabulation, fabrication, or a lie.

2 Likes
45

Ohh noo not an expert… Everyone here is super competent!!! (PS still havent gotten to ur book yet sorry!!)

Oh yes yes YES! Fantastic thoughts! Ye text search is kinda “solved” I guess. You can use advanced data structures like tries, trays etc which solves the text search problem efficiently. 16. Strings - YouTube

Obviously some use a combination of tree methods and hash tables / arrays which speeds up search even more.

Another obvious way is use Count Vectorization. Ie u preprocess text and split it into tokens, and store the sparse array of boolean values YES/NO. The issue is this fails for say searching if a sentence exists in a corpus. THis only works for 1 word. You can try N-grams, but its inefficient.

For memories - this is the holy grail!! The issue as u mentioned is we need to somehow “retrieve” old memories, “summarize/compress” it, and “update” the memory state.

I did mention random ideas in A theoretical novel writing tool - #20 by danielhanchen which I “tried” to show some ideas on keeping a running memory for Transformer based models.

2 Likes
46

Yes very interesting point on the clash between what the system thinks is true vs some new information.

A fine tuned model sounds like a good path yes I agree. We’ll need some new model structure which takes into account long long long infinite sequences.

Current transformers take 2048 pieces. We can enlarge it to say 16K, but we need more context, say 100K or more.

1 Like
47

In mathematics, there’s a system where you can prove certain scenarios given finite information using what’s called induction. Also, there’s Calculus which you can also use to solve for infinity in some situations.

Proof by inductions - Zach Star

I personally have an interest in training a model to derive mathematical equations based on its understanding of our current physical reality. Not only that, but I’d like us to be able to teach the model to apply all the rules of physics it knoews so far in order to help find data in big datasets that statistically prove or disprove theories wherever they lay!

2 Likes
48

Oh I think ur link doesnt work! What does mathematical induction really look like? - YouTube This one right?

Oh yep yep!!! We did induction wayy back in high school, for example using it to prove that a sum of a sequence is equal to some formula given.

Oh yep!! Do you mean like taking limits for calculus? But ye maths has SOOO many cool tools!!

In fact the most fascinating for me is in Quantum Physics - renormalization The Biggest Ideas in the Universe | 11. Renormalization - YouTube. In QM, sometimes adding up all the contributions of every possibility of every Fenymann diagram weirdly adds over 1 (ie infinity). With renormalization, the sum weirdly becomes 1 again. I haven’t watched the video too intently - but now since it reminded me I’m gonna rewatch it!!

You might be fascinating by Dynamic Mode Decomposition, SINDY and other methods for deriving physcial equations from natural systems! Dynamic Mode Decomposition (Overview) - YouTube

But ye if somehow we can augment transformers with say ideas from physics, that’ll be so cool! As you mentioned proving / disproving / finding new equations would be soo cool!

2 Likes
49

That’s the link! Thanks @danielhanchen, I fixed my link btw.

I love mathematics, because like you said, there are so many cool tools! I know the answer is out there somewhere, but how do we use everything around us to figure out what it is? Math has been the easiest thing for me to grasp after understanding that when you put numbers into an equation, they’re simply transformed into something else!

2 Likes
50

I agree! Yep the good ol argument of “is maths discovered / made?”

Even if the formulas for natural systems might look ugly, its quite clear some formula exists which governs the system!

Yep exactly!! Math is super useful and extremely powerful. It’s always so fascinating and wonderful how a few variables mixed in some werid way can give rise to complex world systems!

1 Like
51

I finally made it back to the gym and it feels great :+1: it’s amazing how much wear COVID-19 created by being isolated in our homes :thinking:

1 Like
52

This is really cool @danielhanchen :clap:

Sharing it in this thread to do my part!

Awesome work, really looking forward to it :raised_hands:

1 Like
53

Thanks!!! Super appreciate your support!!! I’ll definitely check the Twitter feed out!! :slight_smile:

1 Like
54

everything in the world is connected together and correspond to each other in some certain way, so we called butterfly effects, but it is so complicated beyond the understanding of human in most situations, using the algorithm and AI as a tool to figure out how the world work and correspond is valuable, but all in all, your app have to been tested in real-world, if it works well, it is have bright future and huge business value. but I doubt prediction in the real world is the strength of gpt-3. do you have any successful examples of your app? if it works well, I could help you to promote in the Chinese market and find investors.

1 Like
55

Hello @mingleejiang!

Oh we’re not using GPT3 for world predictions - we’re making the models ourselves :slight_smile: GPT3 is used for JARVIS - ie you can ask ANY question you like to Moonshot Earth ie “how long will I live”, “how will climate change affect me” or “when will the market next crash”. GPT3 or GPTx or any Transformer model will just be used to parse the questions.

I exactly agree! The world is connected, and yes any small little change anywhere in the world will cause a butterfly effect that changes the future. Chaos theory and dynamical systems models take this into account, where models understand that small changes in systems can affect the final outcome. For example, ramping up the Reynolds number in Navier Stokes for fluid dynamics can cause huge unpredictability.

Hence our gameplan is:

  1. Predict stock market crashes, stocks. Clearly if you can’t predict the market, how the hell can you predict the world? We’re already seeing a 2x returns just by predicting the Dow Jones index. We’re maybe releasing this December 2021 or early 2022.
  2. Predict just Australian (since I’m from Australia :slight_smile: ) societal dynamics - ie life expectancy, disease rates, air pollution, traffic congestion, climate change impacts, house prices - u name it! Moonshot Earth v1 is going to be released maybe mid next year.
  3. Global expansion - US, China, UK, Europe, everywhere! This might be a 20-30 year mission.

If you’re interested in learning more about the technical details: Moonshot Technical Demo + What we've been up to! - YouTube and also our website: https://www.moonshotai.org/

But ye thanks @mingleejiang for your support and help!! :slight_smile: We’re currently chatting away with some VCs in Australia, so I’ll post updates on it here!! We’re launching mid 2022, so can’t show u anything yet :slight_smile:

I can only show u the fast algorithms which will underpin the Earth simulation which has 1.2K Github stars, with us being cited in Microsoft, UW, Greece research papers and algos are now inside NVIDIA, Facebook’s Pytorch and more. We also got contacted by Australia Defence on using our algos but declined due to ethics.

2 Likes
56

Also NVIDIA just announced they’re building Earth-2 - a digital twin of planet Earth to predict climate change!

2 Likes
57

your reply is so detailed and inspired, it is great, hope to have the opportunity to try your app in the future. thanks.

2 Likes