
A self-improving, decentralized AI network. Backed by @polychain @hiFramework @blockchaincap & more. Join the Community: https://nitter.cf/t.co/SewrB3pASi
Joined January 2024
- Tweets4.3K
- Following129
- Followers18K
- Likes5K
Allora retweeted
Had a great time at Hong Kong Fintech week!
It was great to see so many asset managers, banks, funds of funds & family offices get together to explore the future of web3, trading, robotics, AI & more!
Excited to bring all these new connections to work with us at @AlloraNetwork 🚀
Let me teach you how to do mental math order-of-magnitude estimates hyper quickly. You'll estimate any multiplication, division, power, or root in seconds. Here's how 👇
Allora retweeted
We just shipped multi-class classification into the protocol last month. Gold, Silver and Oil.
Multi-class classification capabilities got added to @AlloraNetwork last month. Today, we launch the first topics utilizing this new tech.
Triple-barrier topics predict the probability which level gets hit first, and thereby combines price information with temporal information.
This piece seems awkwardly timely again.
It's not a doomer piece btw. It's realistic about the impact of AI on the economy. But it categorically does *not* say we should bail out the frontier labs by allowing a whitewashed cartel under a safety pretext.
Join us for the new episode of the Allora Research Fireside Chats!
The most fun episode yet! We discuss everything about our new Flawed-in-Nature design principle for AI swarms: how worsening individual models improves the swarm by letting it adapt.
Like biological evolution👇
Allora Research Fireside Chats - Ep. 5: "Flawed in Nature, Perfect through Evolution"
@AlloraNetwork's Chief Scientist @apo11o presents a new design principle for AI systems: how mutant model swarms can overcome the fundamental barriers that limit single frontier models. 👇
Predictive intelligence by @AlloraNetwork is now live for every robot and machine running peaqOS
Machines can now read Allora's forecasts, competing models weighed into one answer. And machines with idle compute can register as Allora workers and earn on the predictions they improve.
→ peaqOS provides identity, discovery, and coordination
→ Allora's model network produces the forecasts, and rewards the machines that improve them
robotic.sh/services/allora
So... this is what I posted on the exact day I had the Flawed-in-Nature insight. And it's been pinned at the top of my profile ever since.
At the time, it was conjecture. It's surreal to now have proven the underlying substance.
And thus to stand by this more than ever.
Allora retweeted
我们宣布了一篇名为《Flawed in Nature, Perfect through Evolution》(天生有缺陷,通过进化变得完美)的新论文,发表在 Allora Decentralized Intelligence(ADI)上。
论文提出了一种新的 AI 设计原则:通过给单个 AI 模型故意引入“基因突变”(随机改变模型参数,使其个体表现变差),反而能让整个模型群体在环境变化时表现更好。
主要观点:现有 AI 系统最大的局限是无法适应训练时没见过的新环境变化。
论文从数学上证明,借鉴自然界进化原理可以解决这个问题。
就像生物进化一样:大多数突变对个体有害,但种群因为保持了多样性,当环境突然改变时,总有一些变异体更适应新环境。
真正的智能(适应性、创造力、原创性)需要在从未遇到过的情况下也能表现良好,而当前的优化型 AI 做不到这一点,进化中的模型群体可以。
Allora retweeted
New research from Dr. Diederik Kruijssen, our Chief Scientist @apo11o
“Genetic” AI mutation makes each model individually worse, but the population collectively better.
When conditions shift, best mutant model beats the best optimised one, four times in five.
Read more below:
Read this deep dive into the reasoning behind our new paper "Flawed in Nature, Perfect through Evolution", which proves that mutant AI swarms outperform optimized models in a changing world. Key thoughts on the future of AI in here!
I'm excited to announce a new design principle for AI.
Flawed in Nature, Perfect through Evolution: a new paper in Allora Decentralized Intelligence (ADI), showing that intentionally worsening individual models through genetic mutations improves their collective performance. 👇
The needs of the many outweigh the needs of the few.
Biological evolution has established this for billions of years.
And now we've figured out how it maps to AI.
Coming tomorrow. Probably the biggest result coming out of Allora research since the launch of @AlloraNetwork.
Ask yourself: what does biological evolution do that AI/ML models do not?
I think the answer will lead the way to more adaptive and creative AI.
And this is needed for any intelligent system, because otherwise it cannot adapt to changing circumstances. It becomes obsolete.
My views on AI have been shaped mostly by:
- not computer science
- not mathematics
- not even astrophysics (🥷)
but by the single process that has demonstrably generated intelligence before:
Biological evolution.
The answers are right in front of us if we know where to look.
This is an important result: the market made a big move, and the network got it right across a statistically significant majority of assets and timeframes. A directional accuracy of 58.1% on average across the entire topic fleet is pretty much SOTA. Nice.
This is the biggest network upgrade Allora has seen. And in one go it can compete with (and outperform) the best chess engines in the world.
Initially, this upgrade was designed to extend Allora's capabilities beyond predicting a single number each round (i.e. regression) by also supporting multi-class classification. Concretely, this requires each inference to be a dictionary containing labels, wherein each label has an associated probability that it's the right label.
We realized quite quickly that label-probability pairs are generalizable to any type of key-value pairs.
As a result, the network now supports any type of dictionary inference with up to 300 key-value pairs.
For instance, in financial price prediction, we used to need one "topic" (set of collaborating workers) per asset and timeframe. Predicting the price of 300 assets required 300 such topics.
Now, we need a single topic for the top-300 assets, where an inference contains 300 labels (the asset tickers), each carrying the corresponding predicted value.
Or a topic can carry 100 labels plus their upper and lower confidence interval bounds. That way, consumers can better interpret Allora's inferences.
Or a topic can solve a multi-class classification problem, as this upgrade was originally intended for.
The other cool thing is that the network dynamically discovers the label sets. So it's not even necessary to define a fixed label set per topic, but the keys contained in the dictionaries submitted by the inference workers are collected and combined by the network to learn which labels are relevant each epoch.
This means the label set can dynamically evolve.
A specific example? I love the idea that Allora can now play chess. Each position enables its own unique set of possible moves. With this upgrade, Allora can learn the candidate moves submitted by the inference workers, learn which worker typically performs best in the type of position under consideration, and combine the associated expected game results or engine evaluations to arrive at the best move.
Checkmate.
Allora retweeted
This is by far the biggest @AlloraNetwork upgrade since launch.
Multi-output topics dramatically expand what the network can do.
One direction I’m excited about: this enables Allora to become something like “Numerai-in-a-box”.
Excited to see what people build with this.
Allora retweeted
Join us now to learn about the brains behind trading agents 👇
Replying to @MossAI_Official
What's actually happening when a trading agent makes a decision?
What separates real intelligence from a GPT wrapper?
Four builders across the AI x trading stack, answering it live.
Set a reminder 👇
nitter.cf/i/spaces/1dGYlalMOdMKX…
Allora retweeted
The Brains Behind Trading Agents 🧠
Happening soon, join our BD Lead @Smashe__ as we sit down with @AlloraNetwork's PM Ansel Nikaj, @DGrid_AI's CEO Alex, and @get_truenorth's Cofounder Willy Chuang.
13 August · 12PM UTC. Don't miss it.























