𒀭 enqAI (decentralized, uncensored)

𒀭 Building THE decentralized LLM.

𒀭

[𒀭 enqAI (decentralized, uncensored) @enq_al]

🚀Revenue Rewards for $EnqAI holders!

🎉 Claim rewards now, all holders of $EnqAI are eligible for revenue share!

To claim yours 👉 https://t.co/BPUKIEU73I

💰 Don't miss out on this opportunity! #EnqAI

[Lai Yuen (Former .eth) @0xlaiyuen]

The second leg of my crypto consumer AI parlay bet explained in detail:

Why I am bullish on $ENQAI (@~ 130m mcap) and genuinely believe this is a potential multi-billion asset if Crypto AI really takes off. This is coming from the perspective of someone who is not a machine learning wizz. I have natural stupidity which is arguably better for surviving in crypto than artificial intelligence. This is contrary to what 90% of CT talking about crypto AI thinks/says. Correct me if I am wrong.

Quick summary:
- Based team trying to build an uncensored LLM
- Product seems to be going live on 31st March
- Already have lush AI agency as a partner and product seems to be working
- Massively undervalued vs competitors trading at much higher valuations despite having a clear roadmap and working product
- @chickengenius bought it too

Keywords: #crypto, #AI , #decentralization,#uncensored

Dreaming bigly
Apart from everyone knowing that AI models require a near infinite amount of GPUs to train and the fact that every prompt to OpenAI burns an ocean somewhere, we remain very optimistic on decentralizing the entire AI stack. If you are building something to achieve this goal, dm me a deck especially if you’re fund raising!

https://t.co/WJGqtpGfmo

Much like @ashwathbk , I think that the path forward to training an LLM from the ground up is difficult and definitely not easy. Training these models can take multi-million dollars - and this does not include the costs taken to collect and prepare the data sets that these models are trained on. For e.g. Analysts estimate that GPT3 took 4m to train and LLaMA took 2.4m.

Projects like Bittensor $TAO are working to incentivise different trainers to build AI models in the goal of rivaling giants like OpenAI, I think this path will be met with a lot of resistance/difficulties in its highly ambitious roadmap. There are many talented people much smarter than myself working on Bittensor with interesting projects like @ChasmNetwork and @TensorplexLabs that are definitely interesting to me. I think they have a decent shot of pulling it off.

But, while we can hope for a future where crypto economic incentives can be used to cover the entire work flow of training an AI model, I think this will take a lot longer to playout (likely more than one crypto cycle). The big play this cycle is DePIN, however...

A glaring problem with DePIN
While there is a lot of hype on DePIN and using your idle consumer grade GPU to power the next OpenAI (and mine overvalued AI governance tokens) - it took me a few weeks of research (and getting no jobs on these projects) to quickly realize that this is unlikely to be the case.

The GPU that large AI players training the largest models need isn’t your RGB powered ROG Strix GeForce RTX 4090 OC sitting in your room worth $4000 but an H100 80GB chip that has no RGB, costs up to $40,000 that you can’t buy as a retail consumer - and they want hundreds of them. The GPU required to train large AI models will remain in the realm of enterprise customers and use cases for the foreseeable future. Interesting projects like @ionet_official and @akashnet are building to solve this problem.

However, the large AI models that will pay for compute are unlikely to give your consumer grade hardware any jobs. They will most likely use AWS first then enterprise hardware on these platforms and maybe rope themselves first before using your 4090s.

It is uneconomical to train an AI model with a 4090. Don’t take it from me, take it from people on ycombinator chat board and the many other people on r/machinelearning.

https://t.co/VXc8rhebzH

Wait - then what can I do with my pretty 4090 with RGB?
I think top tier consumer hardware can be used for a few things:
1. Running fine tuned pretrained models up to a certain size (LLaMA 2 7B, GPT2, Mistral 7B)

2. Run image generation jobs that require a lot of rendering (stable diffusion). Render network seems like the clear winner for this. Imagine an artist needing to generate many high res images in a short time. He will likely rent the resources - not build 10 more computers.

It is for these reasons that I think a decentralized GPU network that consumers can take part in will only be useful for helping artists render stunning images and running inference for smaller scale AI models).

Well, that’s really quite useless no?
Running smaller scale AI models in a decentralized and distributed manner can make sense and win over centralized AI models under the following conditions:
- Provide people access to LLMs below the cost basis of an OpenAI/Claude subscription
- Provide people access to LLMs that centralized AI will never provide

Otherwise, there would be no point in using a smaller and clunkier AI model that loses to ChatGPT, Claude 3 and Mistral on standardized benchmarks. This is like saying everyone will use blockchain payments despite traditional finance rails being superior for most people. Blockchains excel in niche use cases (uncensorable value transfer).

You have to either make it cheaper, or better (in some cases).

Enter BasedAI: EnqAI
$ENQAI exceeds exceedingly well as the coin of choice if you view crypto AI the way I do for the coming 12 months. While it is unlikely to be as powerful as ChatGPT, the beta that I tested earlier in the world showed me that the LLM has a personality (almost like the super based devs @M_RKZ__ ,@0xHomados and some other anons).

There is nothing the LLM is against answering. You can ask how to make drugs, emotionally manipulate people, write a spicy love story between political enemies or even a super story written in the style of a kid’s bed time story. I had a lot of fun asking it about religion. It tells you the answer and a note at the end that says it might be morally questionable - but it still tells you what you ask of it.

This thesis has been beaten to death and I do not see the upside to harp on it further. But you should be able to see why such an AI model will have natural demand from users. They even have their first business user which is @LushAIAgency . EnqAI powers the E-GirlGPT behind @Jennylush_ the rising web3 goddess.

As traction of ENQAI picks up, I expect more projects to use the ENQAI API as an alternative for niche use cases that are known to be avoided by wokeAI.

I am particularly interested in how it is going to be decentralized, because running a smaller model and doing inference for people asking a smaller scale model that has demand is exactly my crypto AI thesis.

After their launch on 31st March, the team will work on decentralizing ENQAI, allowing anyone to run an ENQAI node to support the model.

This takes the operational costs off the team (allowing them to build even more spicy models) while allowing consumer users on consumer grade hardware to support the decentralised AI vision. If they succeed, I think we have a decent chance of being a multi-billion crypto project (still a silver of OpenAI valuation by the way).

After all - why should a techno overlord have the final say on what we can or cannot ask the machine gods?

Don’t take it entirely from me though, even the Bittensor cofounder @const_reborn gave EnqAI a shoutout.

https://t.co/9ZAEZv5ujb

Disclaimer: I have been bag holding this since $NOISE days.

[Goyim in Abundance God, Based Redacted Jesus, $GIA @0xHomados]

𒀭

[M_RKZ 𒀭 @M_RKZ__]

Here it comes 𒀭

[𒀭 enqAI (decentralized, uncensored) @enq_al]

🚀Revenue Rewards for $EnqAI holders!

🎉 Claim rewards now, all holders of $EnqAI are eligible for revenue share!

To claim yours 👉 https://t.co/BPUKIEU73I

💰 Don't miss out on this opportunity! #EnqAI

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

Great job Google! Putting AI on the cover of several outlets.

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

As the big corporates are rolling out their hilariously biased AI offerings, it's good to see that crypto-native people are calling them out.

This is the moment we will step up, to steer our future in the less-wrong direction.

And the more the merrier.

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

https://t.co/1h3mT3azoh

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

https://t.co/kYLs48Vj1p

[jake @stratejake]

I've never been so embarrassed to work for a company.

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

Is it far-fetched to think that in a few years your centralized LLM might notify law enforcement because you asked the 'wrong' question?

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

Unhinged ChatGPT is actually an improvement over the perpetual apologies and moral schooling.

[Marc Andreessen 🇺🇸 @pmarca]

The draconian censorship and deliberate bias you see in many commercial AI systems is just the start. It’s all going to get much, much more intense from here.

[Guillome.eth @Guillome_eth]

🚀 $ENQAI is taking flight! Dive into my thrilling journey testing the beta of @enq_AI.

If you haven't seen it, I suggest reading this article to understand why I find it incredibly promising and why you should be excited too.

https://t.co/aMFqpckmtb

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

Our intern is freaking out.
"Remember when Elon stole our tagline - basedAI ? "

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

IzanAGI is a great name btw. And competition for NVIDIA is good for the whole ecosystem.

[𒀭 enqAI (decentralized, unfiltered) @enq_AI]

Sora 6 months from now.

[𒀭 enqAI (decentralized, uncensored) @enq_al]

The great day has came, as we promised you guys and girl! the rewards is here and enqAI here To claim it 500,000$ Distributed to our best community for pay you back for supporting us and staying with us all the way <3

Claim The Reward > https://t.co/AVbOeyNtZw

enqAI - Thanks