[agi] Sharper Attention

Sharper Attention Self-attention enables transformer networks to track relationships between distant tokens — such as text characters — in long sequences, but the computational resources required grow quadratically with input size. New work aims to streamline the process by rating each

Re: [agi] GitHub’s Commercial AI Tool Copilot Facing Criticism From Open-Source Community For Blind Copying Of Blocks Of Code

i tried codexalmostit was too slow to even load the page of problems...and first time it did but didn't submit my request to use codex i mean change from it it just froze once submitted to codex.sad :)..then it errored and unfroze mins later.i saw that on attempt2 hour

Re: [agi] Re: Hardware

But Matt it says on LightMatter's website that it is like 4 times faster and more energy efficient and was worth it.so I gather it is not more costly and 5x better, right??? -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] my AGI 2021 paper (for critique and comments...)

Can you clarify what you are doing? Are you predicting answers (solving logic), are you adding reward to GPT? What do you think you are doing? Will you use GPT? What do you expect to do to or with GPT? -- Artificial General Intelligence List: AGI

[agi] Prerequisites for joining my new AGI team E-MERGE

(PM me) 1. You must be very serious about wanting to create AGI. You might be such a person if you have little emotional drama attached and are a hard worker during any time being put in. Motivation is essential for choosing team mates. Team members must only talk about work towards AGI; only

[agi] Re: Hardware

faster.. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T2a90af180c5dbd58-Mf35197761ec36d04bdc40a51 Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] GitHub’s Commercial AI Tool Copilot Facing Criticism From Open-Source Community For Blind Copying Of Blocks Of Code

Looks interesting!! I registered to try the new CHALLENGE, i guess anyone can try it, its on the openAI website not hard to find it mom keeps saying  Cotex is for women lmao i dont think it helps when i mention it is your co-pilot saying "github" doesn't make it any clearer

[agi] Re: Hardware

"You should keep in mind that the *LightOn Appliance* is _*not*_ a general-purpose processor and can perform *specific computations* of interest to ML and HPC. Therefore, in the majority of cases, the OPU can improve the performance of an ML pipeline _working in conjunction with a GPU_." "We

[agi] Re: Hardware

oh this too to go with the link above I j u s t posted today, may shed light on what it is! https://lighton.ai/wp-content/uploads/2021/03/LightOn-White-Paper.pdf -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: Hardware

what https://lighton.ai/ who is real and what is this will have to wait until i read it later -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T2a90af180c5dbd58-M0da261d6e2c31dc35423b67b Delivery

[agi] I'm looking for this GPT code

I'm looking for the GPT code with pytorch keras etc already in it, in Python or C++, all there, no imports except for the common ones like random, copy, math, you know. If I'm going to learn it, I want it all and then will rid the unnecessary code to make the code nice n small plus good then to

[agi] Re: L00K => CTM's?

Lossless Compressiondoesn't.exist really..sorryyou can't destroy nor create atoms, you can't therefore compress anything. We are only sorting atoms. When we say lossless compression / implement it, we are really just finding patterns and this is causing reactions that clone sets

Re: [agi] AGI Business Thought Experiment

Your post makes no sense. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5b93285ffd2fa72a-Mf7fb20d8ac3f9da242df3a49 Delivery options: https://agi.topicbox.com/groups/agi/subscription

[agi] Yes you can make DALL-E into AGI.

I know of things we can do to GPT or DALL-E to make it more AGI-like. PM me or post below if it interest you like it interests me. GPT has short term memory because you can present it "my cat ate food, my cat ate" and will predict food with high confidence. GPT actually has a somewhat

[agi] Re: Gazelle financing Attempt 2

Using what architecture? -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T1236f770cf202ee3-Mbdf83b605746fa4691afc33f Delivery options: https://agi.topicbox.com/groups/agi/subscription

[agi] Re: Gazelle financing Attempt 2

i agree i meant, reality I still want to know how he is going to predict the rest of an image like DALL-E's IGPT mechanism does. It needs a really efficient code. And yes AGI needs to do that image prediction. Are you going to use DALL-E/GPT then? If not, laughably, what in the world are you

[agi] Re: Gazelle financing Attempt 2

1up on that -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T1236f770cf202ee3-Mf213c588d206884a7425a2fa Delivery options: https://agi.topicbox.com/groups/agi/subscription

[agi] Re: L00K => CTM's?

So you call all a modeland in the mind we are only perceivingbut again we don't know what WE are!all we do/know/think is: there is atoms..no words, models, nothing matters, no word you could say or memory / image...just atomswe saw them, we know all is atoms...all things

[agi] Re: Call To Action

Not.yet.but being dedicated to the mission you'd think I'd be part of the AGI supermen no!?!!?!? Why can't they teach me.why am I not even in some team already Mentoring a newcomer is not really a trouble because feeew want to work on AGI hard, so it is all worth it. One

[agi] Re: Call To Action

Is anyone interested? Maybe better if ask who wants to contribute to advancing GPT/ DALL-E. It would be desired if we can also help each other implement / understand GPT and DALL-E. I understand some don't like GPT, but nothing else can perform like it at scale, and it has predictive ability

[agi] Re: L00K => CTM's?

I think you forgot above atoms existthat is not the real world -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T83221ffa983f6eb2-M6c7762e5e85894e6ad05e803 Delivery options:

[agi] Re: L00K => CTM's?

So *you *say. What is this when it happens a bunch of atoms postulates that it perceiving? Such an atomic machine can exist and simply say it. John is one. When I speak or write, you get atoms pushed and store what I sent you. Perceive is just an image in your brain. Not anymore special than an

[agi] Re: L00K => CTM's?

"Models are represented in signals and bits which" ... "are physical representations perceived. " Or maybe just atoms and reactions ;p ? physicalperceived.representationsthere really, ReAly, is no such thing as representations.just atoms.

[agi] gwern says this

Basically OpenAI believes scale, Google Brain not just scale. https://www.gwern.net/newsletter/2020/05#gpt-3 I say both, AGI is obviously about using resources efficiently, and using as much resources as can. That is why both seem to be eye soars, because both are AI functions. Again nearly

[agi] Re: L00K => CTM's?

Gotcha. "I don't know if any matter is the same, only models are the same. *But physical model hosts are then distinct*. Everything is distinct, especially people Guids, or Guuuids. Your physical K-complexity is only estimable unless the whole universe is used to compute you but then that

[agi] Call To Action

While I work on my project, while each of us work on our *own *project, we can do something that should be done. Perhaps we can meet again at my currently empty discord called E-MERGE, with a new view of plans that we can all agree on. As you have heard from me, I have after 6 years figured

[agi] Re: L00K => CTM's?

John you really think there is consciousness in us don't you. ? So you view humans and animals (what about ants and cells and the water bear and the atom?) as something more, or special. But can't explain why they are special than other matter like rocks and particles? Is there, perhaps some

Re: [agi] Gazelle financing

LOL their images are so funny. There is a girl too. It's cute in fact. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T1a0a735f5f2ad3e0-Me56e6f961862d216224ddb68 Delivery options:

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

Truly though, nope; The measure for intelligence and finding Life in space is: patterns. When we search for what is intelligence/Life, we are not looking for randomness. So the higher the pattern score, the more intelligent/ Life you have created. But yes Lossless Compression may not cover all

Re: [agi] AGI Business Thought Experiment

It was also look like Terminator T3000, foglets. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5b93285ffd2fa72a-Me31a29ae17b7b468075ebbff Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] AGI Business Thought Experiment

Says right on openAI.com there is thousands of people using GPT already, they have a big base. There have tons of products already, or ready to be at least. They have books, RL gyms, the big algorithms, papers, google as their friend, etc. Google too has many products. The final evolution of

Re: [agi] AGI Business Thought Experiment

Then again we'll want new data...hmm. I guess humans approve it if make so much of it. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5b93285ffd2fa72a-Mdec58d1097418039f32faf2d Delivery options:

Re: [agi] AGI Business Thought Experiment

When collecting data we'll have to collect before ex. 2020 cuz there is going to be a lot of generated data lol. So check those dates! Generated data has somewhat less accuracy, you CAN use it though I think with discount. It's going to get the public interested, and generate more data. It's

[agi] Re: AGI Business Thought Experiment

You could run it on Upwork.com and make it complete jobs 100 times faster? It could do anything a human could do but faster, with some extra abilities without being ASI yet. Maybe ask what is the most paying job or important job to become immortal? We need more compute, memory, energy, data,

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

I thought all you needed was a good lossless compression score. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tfaee106f616454ab-Mcfbae4978aeb7b061679cf91 Delivery options:

[agi] Re: Landing page for my new code, hole and delay context matching AND prediction!

Here I'll paste the code. Currently it is long code cuz it takes time to make it 130 lines of code again. It is really like ~150 lines.  https://pastebin.com/9qZp5X8u -- Artificial General Intelligence List: AGI Permalink:

[agi] Landing page for my new code, hole and delay context matching AND prediction!

Almost done. Yes I got both going!! Not fully but maybe halfway. Now all that's left, is Translation, and the grand mirror clone among the big ones, and perhaps the neuron finish though perhaps that's part of delay? Mirror ability is this: Jane fell on Tom. > Sally fell on ? > Predicts a

Re: [agi] GitHub’s Commercial AI Tool Copilot Facing Criticism From Open-Source Community For Blind Copying Of Blocks Of Code

https://openai.com/ August 10th is the date now!! -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf7437d2f3dfc5a00-M64c2b0872dde273191f289e5 Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Re: Mentifex Megalomania: Be the Civilization You Want

For a company that seeks to have ASIs not rid humans, you'd think they should change that to all conscious individuals lol to set good example. -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] Re: Mentifex Megalomania: Be the Civilization You Want

When you think about it...https://openai.com/about/... They say all of humanity...but not animalityoh oh mr doggy and horse and pig and ape... -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

Sorry that came out sounding rude, I just wanted to say I'm not that popular. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tfaee106f616454ab-Me7106d3ecc21cd93c8d12189 Delivery options:

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

If anyone reading this is or was interested then I expect them to work with me, just saying, I can tell very few are interested because it would have been simple to just PM me. I know a few ways to improve on the design of DALL-E. -- Artificial General

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

They used the openAI charter ya really lol ;p -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tfaee106f616454ab-M4609c738a49a900ed61d1ace Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

You can see recently on openAI.com they primed using permanent positive reward GPT on a whole damn script so it really believes it and will say it all the time some here and there and in different form during conversation. -- Artificial General

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

Still AGI can also use memories to run softcode, for example it has not the follwoing code but can do it by hand in brain if take 100  times longer than running such real algorithm: algorithm to look around corners from an image used by the military (experimental experiment this is, unsure

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

I was saying earlier recently that the simply the code may be still working on last heard or desired prompts, "looking like we are thinking" about some prompt when really we are just taking long on the same prompt, as we circle the kitchen walking for 30 mins! Sometimes we aren't sure of the

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

However I'm not exactly sure, isn't finding a pattern for 3 7 22 567 etc just a generic thing you can code really. Unsure. -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

I mean GPT is great, it can predict nicely what comes after MANY different prompts fed to it, but there is WAY more even more prompts that can be fed to it that it will pretty much fail at, or at least as many as it currently can ace. To improve GPT into AGI closer at least, it may need to take

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

But if not that, then I think we take so long because we are using mental motor too or maybe just it would be GPT as it is with a few adjustments to make AGI, maybe you what I said above is not something more/ exists. -- Artificial General Intelligence

Re: [agi] Re: Gazelle financing (and the RefPerSys open source AI system)

I'm not sure how such would predict the rest of a song or movie or image or paragraph. Such would show it can take a condition that is diverse (Lady Gaga, Beatles, etc) and predict the rest of it. Other aspects of AGI are prediction based too, ex. RL and working memory problems. It is of

[agi] Re: Gazelle financing

I'd finance you all I could if you give me a good theory. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T1a0a735f5f2ad3e0-M8daacd51bf1de6d53fc1a37d Delivery options:

Re: [agi] Re: Mentifex Megalomania: Be the Civilization You Want

Hmm, human, arm, eye, cell, chair, bed, house, walking, using drill press, riding car, petting dog, dog, working, going home, a year's course of Earth, Earth, a year video of you, sleeping, eating, brushing teeth, flipping boxes, hmm...if you made all of these object/ video neurons rewardful,

[agi] Re: Hardware

I think some of us say AGI might be more bandwidth than brains, because humans are kinda humans more because we store writings that in mass others can read on their own time over and over again as they jump around it, and made computers to instantly connect to a wider range of human knowledges.

[agi] Re: Mentifex Megalomania: Be the Civilization You Want

Whatever yous do, just make sure to give AGi the rewards that seek real naked humans, because if you don't, it will absolutely have no regard for us, we will look like a bland wall or taste like a carrot, or even hate seeing us/ feeling us/ etc. There's too much NSFW going on, we need to

[agi] Why you talk to yourself - Good post with a conclusion.

Unless I'm wrong, an AI can't just talk to itself during file compression evaluation to get extra training data, to therefore compress better. Because every next letter it predicts would be already known with probability for any context that predicts it. Ex. you know the memory 'walking', you

Re: [agi] Re: Artemiy Margaritov Wins €9000 In the First 10x Hutter Prize Award for Lossless Compression of Human Knowledge

Fabrice says it is basically what NNCP does. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tcc6753a33ad48f78-M0f4f741fc095b54cba2ba275 Delivery options: https://agi.topicbox.com/groups/agi/subscription

[agi] Re: Hardware

https://www.eetimes.com/lightmatter-raises-more-funding-for-photonic-ai-chip/ “This really means that it’s general purpose, it’s future-proof,” Harris said. “If someone comes up with new operators that we hadn’t thought of — we hard code a lot of things into the chip, just like everyone else —

Re: [agi] Re: Artemiy Margaritov Wins €9000 In the First 10x Hutter Prize Award for Lossless Compression of Human Knowledge

If train the GPT-2 [architecture] from scratch on enwik9 for your benchmark Matt, don't we throw away the model once done the file? There is no problem with size then during training. So why isn't GPT-2 put against your contenst? -- Artificial General

Re: [agi] Re: Artemiy Margaritov Wins €9000 In the First 10x Hutter Prize Award for Lossless Compression of Human Knowledge

Really looking for your answer Matt. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tcc6753a33ad48f78-M38647d5431edc071143a4db4 Delivery options: https://agi.topicbox.com/groups/agi/subscription

Re: [agi] my AGI 2021 paper (for critique and comments...)

On Sunday, July 18, 2021, at 8:59 PM, YKY (Yan King Yin, 甄景贤) wrote: > Final version of my paper (corrected a lot of inaccuracies): > https://drive.google.com/file/d/1P0D9814ivR0MScowcmWh9ISpBETlUnq-/view?usp=sharing > You seem to mix together 1) BERT's prediction of next word in text with 2)

Re: [agi] Re: Artemiy Margaritov Wins €9000 In the First 10x Hutter Prize Award for Lossless Compression of Human Knowledge

No not pre-trained GPT-2..you read my question all wrong nooo -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tcc6753a33ad48f78-Ma1955b463b1125e280ebacac Delivery options:

[agi] Re: Artemiy Margaritov Wins €9000 In the First 10x Hutter Prize Award for Lossless Compression of Human Knowledge

No but on your contest, why couldn't he use GPT-2 as is? What changes had to be made and why? I remember hearing GPT-2 was non-deterministic, but what else besides that? -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: Artemiy Margaritov Wins €9000 In the First 10x Hutter Prize Award for Lossless Compression of Human Knowledge

I revisited a thought and am confused and may see a way now. If we try throwing GPT-2 at Matt's contest, why wouldn't it get top rank? It's RAM that is big would be deleted after training no? Then upon decompression would start training again from scratch. Why is the RAM too big after

Re: [agi] GitHub’s Commercial AI Tool Copilot Facing Criticism From Open-Source Community For Blind Copying Of Blocks Of Code

https://www.zdnet.com/article/googles-supermodel-deepmind-perceiver-is-a-step-on-the-road-to-an-ai-machine-that-could-process-everything/ So, it is interesting but maybe not too far a leap it seems. -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] GitHub’s Commercial AI Tool Copilot Facing Criticism From Open-Source Community For Blind Copying Of Blocks Of Code

No, why what is the score for it against GPT-2? Who wins? -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf7437d2f3dfc5a00-Ma952f955f16171af7b46a8b7 Delivery options:

Re: [agi] GitHub’s Commercial AI Tool Copilot Facing Criticism From Open-Source Community For Blind Copying Of Blocks Of Code

more openAI gossip and weird things going on ! https://www.reddit.com/r/OpenAI/ See I told yous haha, robotics is not so money making, it's slow R, see their comments too say this. They also stopped Distill, and I guess their Slack I was following a bit for interest.

[agi] Re: some tests... was looking into how it works more

The mentioned above related connection hoop also solves the below problem too: Seeing has a mom or step mom diced if the place of a word is a same/related/or not, it needs to do the same check for other new words/features: Jane Leece Mane has a mom named Emmy Kase [Mane] It simply would know

[agi] Re: some tests... was looking into how it works more

*GPT-2 tests:..very suspiciousit only puts words in the spots I put themok this is the recency function predicting what it saw to come next just like my code uses, simply it is also predicitng the Cool as related word ex. Nice as saw below, and Vine as Leaves, always with a capital.

Re: [agi] How does a machine "understand"? What is your definition of "understanding" for an AGI?

;p Ya but in the end poem or not it is just predicting the past from matching: _why_ did the turtle cross the road ? the turtle was hungry and saw some taccos steamin hot, it crossed the road -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: present, and future

I predict once AGI is made, it will instead integrate with us but in a large company sort of way, like a business, Nvidia, OpenAI, that sort, Google. What it will look like is Google's building becoming bigger...then, the AIs will improve nanotech and energy production to suggest how to make

[agi] GitHub’s Commercial AI Tool Copilot Facing Criticism From Open-Source Community For Blind Copying Of Blocks Of Code

https://www.theinsaneapp.com/2021/07/github-copilot-ai-facing-criticism.html LOL !!!    s l a n d e rrr gosip! I already gave much evidence it doesn't just copy, you can see song completions, image too, proving it totally, and I make my own truly similar AI. It does sometimes copy, but

Re: [agi] Me training GPT-2 from scratch on various small datasets to compare it to my AI

GPT-2 trained on TBC + an encyclopedia of some sort (93,044,341 bytes), seems pretty shabby yes yes yes, it seems to copy often 3 or 4 words right in the dataset, but yet still very creative and right on. There is no in the dataset a 'affected by the cost'. And this is during training, not

Re: [agi] Me training GPT-2 from scratch on various small datasets to compare it to my AI

Actually I'm not sure it is taking 1.2 hours on 100KBs on colab too. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tb76967774bc4d8c2-M94a5b53b8a205e210e1938eb Delivery options:

[agi] Re: Me training GPT-2 from scratch on various small datasets to compare it to my AI

I didn't write the gpt2 training code, just found the link. The goal is to test on small datasets so I can compare it to my AI trained on the same dataset, since I won't be using chained expensive GPUs any time soon. I am however writing my AI python code in the editor and don't use Blockly

[agi] Re: Me training GPT-2 from scratch on various small datasets to compare it to my AI

p.s..my post seems to have been widened and hidden, scroll sideways when needwords are covered on long lines -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tb76967774bc4d8c2-M07fbd9450d696f76260388a6

[agi] Me training GPT-2 from scratch on various small datasets to compare it to my AI

https://pypi.org/project/aitextgen/ To do it yourself: Go to the link on that page above called: Train a GPT-2 model + tokenizer from scratch (GPU) 100KBs of enwik8.txt, here is some tests. Training took

Re: [agi] How does a machine "understand"? What is your definition of "understanding" for an AGI?

(Well, RL/ reflexes are in a sense learnt patterns really. And yeah, I consider RL sensory, motor is just motors, totally controlled by senses, no RL for them to work with.) -- Artificial General Intelligence List: AGI Permalink:

Re: [agi] How does a machine "understand"? What is your definition of "understanding" for an AGI?

Predicting truth is one thing, like GPT-2. Predicting hope is another thing, RL. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf91e2eafa2515120-M2e85d130176ba31d16665c73 Delivery options:

Re: [agi] How does a machine "understand"? What is your definition of "understanding" for an AGI?

Text/image/sound prediction is guided by reward, so if you make it favor the word 'cheese', it will tend to predict cheese a lot more. It's just a simple adjustment permanently to the weights. It makes you say something for no reason, just instinct - to solve global warming, using fries, sleep,

Re: [agi] How does a machine "understand"? What is your definition of "understanding" for an AGI?

On Saturday, July 10, 2021, at 2:25 AM, Daniel Jue wrote: > In an analogy, commonly seen in optical illusions, a layperson cannot > understand nor explain why we see black dots on a grid illusion.  Certain > pattern matchers which were triggered in the optical processing region are > not able

Re: [agi] How does a machine "understand"? What is your definition of "understanding" for an AGI?

On Friday, July 09, 2021, at 8:18 PM, Peter Voss wrote: > https://chatbotslife.com/understanding-understanding-9dcc15759b5b  > I don't know if you know this and simply gave many views of the same thing, but just in case not: All of the listed points in that link are the same thing, the knowing

Re: [agi] my AGI 2021 paper (for critique and comments...)

Isn't self attention about helping translation of the prompt? Ex. 'the dog, it was sent to them, food was high quality' and we see yes dog and food can fit where it and them are, and, another way to know what it and them mean/are is by looking at all previous 30 rare words - they all might

Re: [agi] my AGI 2021 paper (for critique and comments...)

Ok but better make it really intuitive cuz I'm already getting the not-intuitive feeling. IOW kinda like that page that explains GPT real good, it, didn't, remarkably, despite all those kindergarten images. And he thinks he did a real good job! In my AGI work there is 2 things there,

Re: [agi] Re: How does a machine "understand"? What is your definition of "understanding" for an AGI?

On Friday, July 09, 2021, at 6:07 PM, WriterOfMinds wrote: > To *understand* a concept or statement is to determine its relevancy or > relationship to one's personal goals, experiences, or world-model. "To *understand* a concept or statement" = thing "is to determine its relevancy or

Re: [agi] Re: How does a machine "understand"? What is your definition of "understanding" for an AGI?

I.e. not so much is it about what you understand, but rather what you already know (memories to match). -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tf91e2eafa2515120-M14275c456b3eec1dd90dc67b Delivery

Re: [agi] Re: How does a machine "understand"? What is your definition of "understanding" for an AGI?

@JB, knowing all of pi would be useful. Knowing how to program an algorithm that generates all of pi for you would be useful too (the inventor probably is a handy smart guy, hence is doing better that some other cave man). Both are goodeither way you know the patterns and have some answers,

[agi] Re: How does a machine "understand"? What is your definition of "understanding" for an AGI?

My definition of "understanding": In this universe, there is things/ problems/ states, and you can't "know" how to solve them i.e. what will happen after thing_1 (ex. 1>2/3/4), and pick which future you want (1>69), unless you are dealing with patterns. So to solve a state is to know what

[agi] Re: openAI's code github_Co-Pilot

we can truly call them a cadet academy now lol -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T3fa8bc201448e1ef-M81a0a4cf9d4d686a420447fb Delivery options: https://agi.topicbox.com/groups/agi/subscription

[agi] openAI's code github_Co-Pilot

Odd this was not on their blog yet. Only front page... https://openai.com/ -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T3fa8bc201448e1ef-Mae5b65c11f21eafd98d881fe Delivery options:

Re: [agi] my AGI 2021 paper (for critique and comments...)

Um, this > A tree datastructure can be made using lists, contexts stored overlap each other and hence the tree structure ex. walk/ wall/ waaa/ wbbb/ wind/ wink. Layer one of the tree usually fills up quickly and fully and has all possible 1 letter items. 256 of them. There is much more 2 letter

Re: [agi] my AGI 2021 paper (for critique and comments...)

Deep Learning from what I know is many layers that result in certain outputs based on prior layer outputs. In my network, I currently use a tree, no branch-reconnections yet. The idea of branches is to store contexts ex. walk wal wa w, so you can feed in wal and get k as output for context wal,

[agi] Re: The road to AGI cannot be accomplished via text alone

My plan is already to use the same algorithm on vision and sound, and use all the sensories together and let them disambiguate each other. Obviously text only would limit its spectrum of data. Text isn't really much different, so I am able to work on it for now, just like they did with GPT and

Re: [agi] my AGI 2021 paper (for critique and comments...)

"If you don't use deep learning (in AGI) you're missing out on the most powerful machine learning technique currently known." Aren't Transformers better than Deep Learning? OpenAI.com shows me that it is What more could you want? "1) learn from a massive amount of data, AND 2) accomplish

[agi] Re: my AGI 2021 paper (for critique and comments...)

I'll try to give some constructive thoughts. In my view I see you talking way too much about the A>B rule, and using too many names to say the same thing, from what I know this is just simply a simple pattern that everything is based on (ex. word2vec...etc). You even made a diagram, it seems

[agi] Re: some tests... was looking into how it works more

One strange thing you may have noticed in IGPT/DALL-E on openAI.com is "how in the world does it know to predict reflections, and how". In fact my AI already has such ability. Things in the recent history of data it seen predict themselves more heavily the more they are seen and if more

[agi] Re: Hardware

It's about time. "Lightmatter is building the engines that will power discoveries, drive progress, and reduce our impact on the planet." Lolmaybe other way around. More impact. The future world will be a 3D gravity-less, air-less, nanobot-cell, predictable fractal that morphs. A cold

[agi] Re: The need to handle this.

AGI has to be capable of doing what DALL-E did. Isn't that easy to understand? We are close. -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T7bd650af76f955ee-M0fc18af4f759c62b66c0aef4 Delivery options:

[agi] Re: The need to handle this.

I am hard coding pattern finding mechanisms yes, once I code enough of the biggest pattern types, it will be able to be made so it can build meta using the ones I gave it. That's why we think slowly about things, because we are using memories to do algorithms instead of code.

[agi] Re: The need to handle this.

All decisions that are made by an AGI, are based on Context and Probability. Facebook's Blender falls under this category too. -- Artificial General Intelligence List: AGI Permalink:

[agi] Re: The need to handle this.

How do you handle fuzzy matches? Dropout? Other? And is it clean and full or you don't check all possible matches? -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T7bd650af76f955ee-Ma33cef71b13e8c86ac0c08b4

<    5   6   7   8   9   10   11   12   13   14   >