Your LLM’s personal trainer. Get with the program, nerds.
- Tweets 232
- Following 11
- Followers 1,555
- Likes 44
Pinned Tweet
The HapticAI beta is right around the corner, and we'd like to take this opportunity to announce our next partner, @heurist_ai!
Heurist is a DePIN protocol for AI inference 👾
Find more details about this partnership at the blog link below or read on for a quick summary 🧵 👇
medium.com/@hapticaidev/hapt…
12
3
24
12,305
we received 30,000,000 $HAI tokens into our treasury today as part of their airdrop.
we have decided to lock this for a year to support the @Haptic_AI team.
etherscan.io/tx/0x68da93c06e…
27
16
92
15,644
Our LBP has concluded.
Over three days, we have raised over $2.5M and sold over 95M tokens.
This wouldn’t be possible without the support of incredible community and the very capable team at @FjordFoundry.
We will be listing on Uniswap in 24 hours!
7
13
77
13,516
At long last.
$HAI is finally live on Uniswap.
Contract Address: 0x89DeB6C8918a42457bD6ddBCAaf979216C4D774C
There are no links in this message, do not click on anything suspicious.
5
11
69
20,421
No seed. No private sale. No advisors.
Just an LBP on @FjordFoundry and a dream.
app.v2.fjordfoundry.com/pool…
5
6
54
18,456
We’re going to be trading on Uniswap later today!
Liquidity will be added on @Uniswap at 10:00 AM UTC today.
Please beware of any scams.
5
8
51
11,862
We’ve been getting a lot of questions in our TG about our airdrop, so he’s some more info:
- Any wallet holding over 250 $MONAI at the time of our snapshot will be eligible for a $HAI airdrop
- The snapshot was taken on the 2nd of April at 1pm UTC
- The airdrop will be proportional to the wallet’s $MONAI holdings
If you want to be a part of the discussion, head over to the link below! 👇
5
7
48
36,609
The following allocations of $HAI tokens have been locked as part of their vesting plans. Please check out our tokenomics for more information.
Team:
etherscan.io/tx/0x88e52ad29e…
Liquidity Provision:
etherscan.io/tx/0x473e93a74f…
Strategic Partners:
etherscan.io/tx/0x1c573716fa…
Incentive Programs:
etherscan.io/tx/0xb9fb792ebb…
4
8
46
4,921
Ownership of the $HAI token contract has been renounced.
etherscan.io/tx/0x1d140147df…
6
4
41
4,460
The $MONAI community can now claim their $HAI airdrops on @TeamFinance_
We urge everyone to be wary of any scams, we have not posted any links to the airdrop.
End of communication.
11
2
44
23,027
gm
We’re excited to announce the launch of our open alpha!
Head over to our website to provide feedback on @monaidev’s inferences. Your feedback is valuable and might be important later! 👀
Join our TG too if you haven’t already to get in on the discussion 👇
7
9
40
6,133
🔥
Honored to be joining the NVIDIA Inception Program. Joining this program will help us scale faster, broaden our network and expand our reach.
monai.dev/news
5
2
36
4,848
Decentralised LLMs make for great comics.
We're looking for users who can get @monaidev to provide the funniest responses of them all, and we're giving away prizes in USDT!
To participate, screenshot your responses on the $HAI website, and tag our account in a tweet.
We will be closed for entries in 5 days.
Happy hunting!
7
4
40
8,945
Our team has been hard at work to get Haptic off the ground.
We’re excited to share a deeper look into our long term vision. Access our technical paper below 👇
hapticai.dev/technical-paper…
1
6
35
11,718
Over $1.5M raised, less than a third of the LBP supply left.
Less than 24 hours to go.
Don’t miss the greatest opportunity to reshape the direction of AI.
app.v2.fjordfoundry.com/pool…
1
4
28
7,442
We're building the first truly decentralised source of reliable inference feedback for AI models.
This thread will explain how feedback is provided on HapticAI and the system's ideals.
Let's get to it 🧵 👇
1
7
29
14,271
A process of constant and consistent learning is essential to the long-term growth of any autonomous agent.
LLMs in their current state are disconnected from the end-user experience.
HapticAI aims to solve this. Here's how 🧵 👇
4
8
26
16,811
We’ve entered the final 6 hours of our LBP.
Over 90% of the LBP tokens have been released.
Get your piece of the pie if you haven’t already!
app.v2.fjordfoundry.com/pool…
2
4
26
12,411
Live now from the HapticAI LBP 🐸
app.v2.fjordfoundry.com/pool…
5
1
17
6,938
As HapticAI scales and integrates with the biggest AI models in the space, it is important to build future-proof systems.
This is why our technical architecture is being seriously revamped from the open alpha.
Here's a breakdown 🧵 👇
3
3
20
2,352
More models, more incentives, and far more data solutions.
The Haptic Beta will be dropping soon, don't get lost in the noise. 🤖
3
18
1,648
While the team works on integrating feedback from the Open Alpha stage, our community ought to understand the Haptic process a little better.
Here's a thread on how Haptic could help you and your AI model perform better. 🧵 👇 (1/x)
3
2
17
1,934
are you gonna sit in your mom's basement playing pokimons or be a gigachad and contribute to the future of AGI
1
3
16
2,136
LBP is LIVE!!
Head over to the link in the next tweet to be a part of the future of AI re-training! 👇
3
3
15
11,690
24 hours in.
30% of the LBP tokens have been sold.
Are you really going to let this slip, anon?
app.v2.fjordfoundry.com/pool…
2
2
12
4,490
Human feedback as a loss function to optimize the model can significantly improve the quality of inferences.
Haptic is being built to ensure the accessibility of these tools and the network of high-quality human feedback providers to all AI developers and teams.
1
2
13
1,667
Hot News & Insights in Crypto [ Alpha issue #114 ]
> Earn DROP Points on @cega_fi with every Cega Deposit
> @Haptic_AI - building the first truly decentralized source of reliable inference feedback for AI models
> Paradigm: Crypto's Biggest Skeptics are EVM-Pilled
> Artificial Superintelligence Alliance with @oceanprotocol, @Fetch_ai and @SingularityNET
> Parcl had made Snapshot - @Parcl
> Clearpool Expands to Avalanche - @ClearpoolFin
> mdLP Pool APY % on @Radpiexyz_io
> Avail’s Light Client Lift-Off challenge is officially live - @AvailProject
> GSR markets granted 'major payment institution' license in Singapore
> Deep Investigation of LRT & Comparison
Want to explore more? ✨
Subscribe to my newsletter to enjoy more
🔹Early projects
🔹Market overviews
🔹Exciting updates
delivered straight to your inbox multiple times a week
Join Alpha Newsletter by clicking on a link in my bio (I am afraid of the Twitter algo)
1
2
12
4,517
In Phase 3, we'll be leveraging hierarchical RL for our complex LLMs with memory, while also using inverse reinforcement learning to enhance our reward models.
Context-based feedback collection will also be introduced in our chatbots and assistants, thanks to embedded feedback.
1
12
1,296
With our beta in its final phase of testing, we thought we’d update where we are in our roadmap.
Turn your notifications to not miss when we launch! 👀
2
3
9
1,201
1/6 After months of testing, feedback, and development, we're thrilled to be launching our beta application, complete with a redesigned website and UI that makes the feedback process more intuitive.
2
3
12
921
This is just the beginning, head over to our website at hapticai.dev to learn more about what we're building.
If you fancy yourself a nerd, read our docs hapticai.gitbook.io/hapticai
1
12
1,196
Ultimately, HapticScore is more than just a metric; it is a catalyst for unparalleled excellence, ensuring that the symbiosis between humans and AI continually advances towards greater understanding and accuracy. Through HapticScore, we are not just teaching AI—we are advancing with it.
1
1
10
2,275
Publicly available datasets without specificity leads to a similar convergence in behaviour, regardless of how sophisticated the internal neural networks of individual LLMs are. There’s a stark need for alternative datasets as well as fine tuning.
Haptic solves this.
2
11
1,689
Our long awaited beta is finally here!
Head over to hapticai.dev to interact with our models from @monaidev and @heurist_ai, provide feedback, and do your part for the future of AI!
5
11
1,400
We will revolutionize the way AI training and optimization are carried out with a focus on a decentralized rewards redistribution model to gather the highest quality of feedback data.
We are committed to providing the nuanced human input that is crucial for refining LLMs.
1
9
1,155
Multi-Agent Reinforcement learning (MARL) is the Future of training the next generation of AI applications
3
4
949
We’re thrilled to announce a powerful new partnership with @ChasmNetwork. This partnership will revolutionize AI deployment across various use cases.
Here's how 🧵 👇
medium.com/@hapticaidev/hapt…
1
1
7
1,428
In Phase 2, we will leverage the data collected in Phase 1 to further refine our LLMs.
This phase of the project is focused on the application of Proximal Policy Optimization (PPO). We will harness this data to fine-tune the models in line with user preferences and expectations.
1
7
347
Importantly, HapticScore maintains a rigorous standard for feedback quality. Contributors whose scores fall below a designated benchmark face a reduction in their Haptic Token stake, a measure to ensure that only constructive feedback influences AI training.
1
1
8
1,628
We will stay open for entries until Monday 5PM UTC!
Decentralised LLMs make for great comics.
We're looking for users who can get @monaidev to provide the funniest responses of them all, and we're giving away prizes in USDT!
To participate, screenshot your responses on the $HAI website, and tag our account in a tweet.
We will be closed for entries in 5 days.
Happy hunting!
3
9
2,260
AI agents are the biggest thing happening in crypto right now. It started with Andy Ayrey running the infinite backrooms and the conversations between AI leading to the creation of Truth Terminal and $GOAT
5
1
8
3,366
The adaptability of HapticScore stands out as its most commendable feature. As the landscape of AI evolves, so too does the benchmark for quality feedback, positioning HapticScore as a metric in constant evolution.
2
1
6
2,108
We’re happy to announce that @pitta_vata won our prompt contest!
This is going to be the first of many earning opportunities on Haptic 💰
Turn your notifications on to make sure you catch the beta launch 👀
2
2
6
1,009
High-quality training data is very important when it comes to developing effective AI models. This involves meticulous data engineering, ensuring the data is accurate and well-prepared leads to reliable model performance.
Here's how Haptic does it 🧵 👇
1
2
6
727
LLM models will compensate Haptic based on a composite value derived from the number of high feedback queries and iterations of retraining their parameter set.
The first phase of our operation focuses on fostering active participation from users as they engage with LLMs.
1
7
295
Synthetic datasets are becoming increasingly important in the AI domain.
But what exactly are they? And how can Haptic integrate and improve them?
Simply put, synthetic data is artificially generated rather than collected from real-world events or sources.
1
2
6
488
We are closed for entries. The winner will be announced tomorrow!
Decentralised LLMs make for great comics.
We're looking for users who can get @monaidev to provide the funniest responses of them all, and we're giving away prizes in USDT!
To participate, screenshot your responses on the $HAI website, and tag our account in a tweet.
We will be closed for entries in 5 days.
Happy hunting!
1
6
1,516
Functionally, HapticScore assesses feedback provided by network contributors against rigorous criteria of precision and relevance. High scores signify impactful feedback, crucial for enhancing AI responses effectively.
1
1
6
1,219
Our primary business is feedback data, so most of the attention will be dedicated to processing and storing this data.
As a user interacts with a model through Haptic and provides feedback, it is broken down and fed to various parts of the system. 🔐
1
6
137
We've been talking a lot about Reinforcement Learning in LLMs, but it's making waves in other areas too. Let's dive into some fascinating non-LLM applications of RL!
1
1
5
811
Let's talk about vector databases? Simply put its a supercharged search engine for AI.
Instead of storing plain text, it stores data as mathematical vectors, allowing for lightning-fast similarity searches.
1
1
5
1,063
This is the last tweet.
Website: hapticai.dev
TG: t.me/hapticai
5
4,738
Finally, the data is presented to projects in a manner that they request.
We pride ourselves on our ability to craft bespoke solutions to convey insights from the data we collect.
Eventually, Haptic will become the black hole for human-trained AI models.
1
6
524
3️⃣ System ideals
Unbiased feedback can only be received from a level playing field.
This is why HapticAI has never sold any tokens to investors or advisors. Having every potential user on the same footing is essential to our success.
1
4
2,352
A novel incentive structure is integral to HapticScore, designed to acknowledge and reward exemplary contributions. Superior scores correlate with higher rewards, promoting a culture of excellence and motivating participants to uphold quality standards.
1
1
5
1,205
👀
Introducing the first LoRA model deployed on Heurist🪄
Magus Devon is a witch who uses enchanted mushrooms to journey to other realities. She traverses between multiverses, and owns a shop where she sells artifacts from her quantum trips in exchange for a good tale.
@MagusDevon (Madotsuki), the creator of this model, is a community manager of @RuniverseGame and the owner of the Forgotten Runes wizard Magus Devon of the Quantum Downs.
Over the past year, Madotsuki has crafted hundreds of LoRAs to bring Magus Devon's adventures to life, capturing the unique look, personality, and quantum magical abilities of the quantum witch.
What is LoRA? LoRA models are like add-ons to Stable Diffusion to make pictures of specific styles, characters, or concepts. They are typically trained with dozens of images. You can find Magus Devon's training image collection for inspiration: drive.google.com/drive/u/0/f…
Visit Heurist Imagine. See the amazing world of Magus Devon, made real by lots of AI training and her creator's love 💜
Learn more about the lore: forgottenrunes.com/lore/wiza…
5
708
After this stage, the valuable feedback provided by legitimate users needs to be processed to be prepared for storage purposes.
We are also working on proprietary tagging and classification algorithms so that projects may derive more insight from Haptic's raw data stores.
1
4
104
This integration will create a continuous stream of high-quality training data, meticulously vetted through human interaction.
Open-source AI models on Heurist's platform will benefit tremendously from this.
1
3
340
Heurist provides the compute infrastructure and support for image/video generation and open-source LLMs.
Haptic’s feedback network can refine data for a diverse set of AI models enabled by Heurist, providing our community opportunities to contribute to a large network of models.
1
3
421
It’s been a while since we announced our website and dApp revamp, and initial feedback on UI from users has been positive.
Before we dive into some exciting updates, a quick run through on how you can use Haptic today.
1
4
561
There is dilution of AI agents category through rapid launches via platforms like @virtuals_io @vvaifudotfun @ai16zdao @SimulacrumIO
2
1
4
948
We're excited to share what we've been cooking with you! Head over to our links below to interact with our community 🔗
TG: t.me/+PuGP8s7oG58zMjA1
Website: hapticai.dev
2
642
The readily available high-quality data from Haptic’s RLHF network will significantly reduce training time for Heurist's upcoming AI models.
Eventually, we see this partnership possibly leading to unlocks in human-centric training techniques applicable to various AI domains.
1
3
760
The Future? Expect to see tighter integration between vectorDBs and models. Hoping to see hybrid systems where line between database and the AI model becomes blurred, leading to more powerful and efficient AI systems. DeAI projects are uniquely positioned to experiment here
1
1
1,156
Once you click the "Submit Feedback" button, a popup should open up. This is the actual interface where you submit feedback on the inferences.
Currently, users are expected to score the inference on a scale of 1-10, with 5 as an average, and justify their score.
2
3
469
HapticScore is predicated on a straightforward yet impactful notion: feedback quality varies significantly. Through meticulous evaluation, HapticScore ensures that only the most accurate and relevant feedback steers the AI's learning trajectory.
1
3
641
2️⃣ Bad scores
Bad inferences happen, and our product is intended to reduce the frequency of their occurrence.
When you do have to provide a low score, you will be asked for additional justification.
2
2
400
An integral part is ensuring that users are real and feedback is valuable.
This is done by the User Integrity and Feedback Quality Engines. These tools can help identify malicious actors who are trying to exploit the system.
The Incentive Engine decides rewards and penalties.
1
4
114
1️⃣ Providing feedback
When you provide a prompt, it is routed to the @monaidev API, where an inference is processed and sent back to your screen.
You then have the option to submit feedback on the inference. This feedback is dissolved into a format the LLM can learn from.
2
3
684
Join us on our journey as we democratise AI and make it more human.
TG: t.me/hapticai
Website: hapticai.dev
Docs: hapticai.gitbook.io/hapticai
LBP: app.v2.fjordfoundry.com/pool…
1
3
3,136
Today, we thought we would help explain key concepts and designs that are critical for under the hood functioning of our network. We will be starting with RLHF. Specifically, comparing two techniques: Direct Preference Learning (DPL) and Recursive Reward Modeling (RRM).
1
1
2
674
Why should you care?
If you interact with any AI models, you may earn through our platform.
Haptic will soon roll out A vs. B scoring, a broad roster of partner models, and a lucrative incentive mechanism for loyal feedback providers.
But wait, there's more! (5/x)
1
3
218
By integrating Reinforcement Learning from Human Feedback (RLHF), Haptic AI focuses on refining LLM responses for improved comprehension and interaction. The linchpin ensuring the utmost quality of feedback? The innovative HapticScore.
1
2
651
Why decentralise?
Openness in AI is critical. We all remember what happened with Gemini.
RLHF without the right incentives can be used to justify bias formation. Majoritarianism is not the intended consequence, but rather the verifiable logic of many guiding inferences. (3/x)
1
3
191
How does Haptic help?
At its core, Haptic is a data company that provides high-quality retraining data to AI models.
Our team builds bespoke solutions for each model based on their unique characteristics so that they may be able to assimilate feedback from our network. (4/x)
1
3
171