The Network for Algorithmic Breakthroughs | Follow the everyday @tigdoesthings https://teapawt.org/t.co/APlJSBonZF
Joined April 2024
- Tweets1.5K
- Following6
- Followers10.7K
- Likes10.6K
The Innovation Game Foundation retweeted
The search for better algorithms shouldn't be limited to people with specialist programming knowledge.
Today, we're open-sourcing Talos, our new autoresearch tool, now available in its initial beta.
Talos takes an idea and turns it into an iterative search process. You pick a TIG challenge and give an LLM a direction to explore. Talos then runs in a loop, writing, modifying and testing algorithms, and comparing each candidate against the current top mainnet algorithm.
Every candidate is compiled and evaluated using TIG's own runtime and verifier. It must outperform the baseline on both training and held-out data to qualify as an improvement.
The result is a new way to explore algorithmic solutions. You use your own compute and LLM account, and get a ready-to-submit package if Talos finds something better.
By lowering the barrier to algorithm development, we can bring more people into the innovation, put more ideas to the test and create more opportunities to find algorithms nobody has thought of yet.
Itβs GPLv3 licensed, so anyone can build on it.
Code: github.com/tig-foundation/taβ¦
Install: pypi.org/project/talos-tig/
Writing a competitive TIG algorithm takes time, specialist knowledge and a lot of implementation work.
We think the search for better algorithms shouldnβt be limited by the few people who have all three, and thatβs part of what weβre working towards at TIG.
The search for better algorithms shouldn't be limited to people with specialist programming knowledge.
Today, we're open-sourcing Talos, our new autoresearch tool, now available in its initial beta.
Talos takes an idea and turns it into an iterative search process. You pick a TIG challenge and give an LLM a direction to explore. Talos then runs in a loop, writing, modifying and testing algorithms, and comparing each candidate against the current top mainnet algorithm.
Every candidate is compiled and evaluated using TIG's own runtime and verifier. It must outperform the baseline on both training and held-out data to qualify as an improvement.
The result is a new way to explore algorithmic solutions. You use your own compute and LLM account, and get a ready-to-submit package if Talos finds something better.
By lowering the barrier to algorithm development, we can bring more people into the innovation, put more ideas to the test and create more opportunities to find algorithms nobody has thought of yet.
Itβs GPLv3 licensed, so anyone can build on it.
Code: github.com/tig-foundation/taβ¦
Install: pypi.org/project/talos-tig/
The Innovation Game Foundation retweeted
Last $TIG community was actually inspiring stuff. Highly recommend to take that time of your day, itll be Well spent.
The Innovation Game Foundation retweeted
Sergey Brin said compute is dessert. The companies winning the AI race right now are not the ones with the most chips. They are the ones with the best algorithms.
Every headline you read about AI is about data centers, Megawatts, and Nvidia orders. Billions in infrastructure and more. The entire investment thesis of the last three years has been built on compute scaling as the primary moat.
Sergey thinks that framing is wrong.
He pulled out an example from physics. The N-body problem. Scientists have been running those simulations since the fifties. Over the decades, raw compute improved on Moore's law. But the algorithms to solve the problem improved faster. Not slightly faster. Far faster. The algorithmic gains made the compute gains look small.
He says the same thing has happened in AI over the last decade.
Compute is not the meal. It is the dessert. You still want it. Nobody is turning down frontier compute. But the companies that figured out the algorithms first are the ones actually ahead.
The market is pricing AI winners by who has the most chips.
Sergey Brin just said that is the wrong scorecard.
The ones who win this are not building the biggest data center. They are solving the harder math problem.
The Innovation Game Foundation retweeted
Thanks to everyone who joined our Community Call last Friday and sent in questions.
We covered the timeline for bringing algorithms into TIG, the recent token movements, mining pool centralisation, upcoming developments and our LLM challenge.
Here are our answers to some of the main questions.
The Innovation Game Foundation retweeted
Tomorrow at 5PM BST, we're back with @Dr_JohnFletcher for a Q&A.
Curious about TIG, our recent developments, or what's coming next? Whether it's something you've been wondering about for a while or something new, this is your chance to get answers directly and hear the latest news from across the ecosystem.
Please post your questions in the TIG Telegram channel ahead of the call.
See you there!
The Innovation Game Foundation retweeted
An AI might find loopholes in the rules we set for it. But those loopholes might not apply in the world running its simulation.
The Innovation Game Foundation retweeted
Hi folks, a quick update from us.
TIG has a busy few months ahead, with key projects and plans moving forward.
So weβre moving our Community Calls to every other Friday, giving the team time to focus on the work and bring more to the table when we meet.
See you at the next one. Itβll be worth it.
The Innovation Game Foundation retweeted
What if a superintelligent AI found a way around every rule we set? Could making it believe it was in a simulation be a way to deter it?
@Dr_JohnFletcher and Adi discuss the Simulation Deterrent.
Read more: lesswrong.com/posts/fdk6rJLkβ¦
The Innovation Game Foundation retweeted
We've now open-sourced the code for a mining pool that runs on TIG.
Want to run your own? Fork it. Give it a name. Set your own rules. Open it to whoever you like.
The pool assigns benchmarks to members' machines, submits their work under the pool's TIG account, tracks credits and collateral on Base, and handles settlements and withdrawals. Most of the groundwork is done. The rest is up to you: hosting, branding, payout model, fees and spending caps.
What's the point? For those new here, pools help smaller benchmarkers balance their factors across challenges. More pools mean more people can take part, and healthier competition across the network.
That's why we're putting the code out there so anyone can build on it. No need to start from scratch.
Pool server: github.com/tig-foundation/tiβ¦
Member client: github.com/tig-foundation/tiβ¦
MIT licensed. Still in development, so expect some bugs.
Catch up on last weekβs Community Call.
Community Call recap!
Can you deter a superintelligent AI by making it believe it might be inside a simulation?
Last Friday, @Dr_JohnFletcher and TIG researcher Adi explored the idea behind the Simulation Deterrent, and why proof of work could play a role.
teapawt.org/i/spaces/1DxLdZnLQVwxmβ¦
The Innovation Game Foundation retweeted
Next up, TIG now has a second X account (hello, it's me).
@tigfoundation will focus on formal announcements from the Swiss foundation, which holds TIG's intellectual property.
@tigdoesthings will share more of the day-to-day activity across the wider ecosystem. Both accounts will help people follow the different sides of TIG.
The Innovation Game Foundation retweeted
$TIG to outperform them all
PRL to outperform TAO.
NEAR to outperform VVV.
ZEC to outperform BTC.
QTC to outperform SUI.
LINK to outperform AVAX.
HYPE to outperform SOL.
LIT to outperform UNI.
Some rough market comparision takes here, it's time to think about outperformance.
Everything will be higher in 3-6 months, but will you beat the market?
I'm planning on it. Let's ride.
The Innovation Game Foundation retweeted
Solid $TIG spaces as usual.
A second mining pool goes open source next week. Anyone can spin one up.
Oxford and MIT experts signing on as challenge owners.
Optimization conference in Munich next month.
Annnnd a proof-of-work case for AI safety I haven't seen anywhere else.
TIG Goes Multidimensional teapawt.org/i/spaces/1DxLdZnLQVwxm
This has been on the cards for a while.
As The Innovation Game grows, so does the work happening around it.
Weβre organising our ecosystem into clearer spaces, each with its own focus. One for the foundation itself, and another for the people, projects, research and ideas taking shape across TIG.
Follow the everyday side @tigdoesthings on X and Instagram.
The Innovation Game Foundation retweeted
$TIG explained in 2 minutes
The Innovation Game Foundation retweeted
"It is a capital mistake to theorise before one has data. One begins to twist facts to suit theories, instead of theories to suit facts."
- Sherlock Holmes
The Innovation Game Foundation retweeted
$TIG community calls are sometimes like a cross between Open University and Tomorrows World .. Gotta be among the best Spaces on X
Come along, have your mind blown by Dr Fletchers ability to predict the future, learn or simply be inspired to question more ..
Tomorrow at 5pm BST, @Dr_JohnFletcher will be joined by TIG Researcher Adi to discuss The Simulation Deterrent.
You'll hear how AI censored publication of the most creative AI safety strategy ever proposed, why itβs in the news, and what this has to do with TIG.
Heβll also cover the latest TIG updates, including the upcoming open-sourcing of the mining pool, our research teamβs trip to Munich, and new Challenges and Challenge Owners.
See you there!
