Continuation for the “Reimagining CLI for AGI”:
Part 4: “Q&A and the Multiverse Computing”
Part 5: “I’m building a Coding Environment for AIs”
Previous videos:
Part 1: Watch this video! :) - #1372 by piluso
Part 2: Watch this video! :) - #1396 by piluso
Part 3: Watch this video! :) - #1413 by piluso
Language is a bit colorful. Worth watching. A very good demonstration of why
the public is ready for Autonomi.
LOUIS!!! COME ON!! Come to autonomi! Are you here?!
Interesting platform for decentralized compute on secret data.
Is this a multi user code for promotion purposes? I’m guessing I can only hand out my code once?
At this time codes can only be used by the first person to use it. The site may accept more than as if it works but when they validate it only the first one will benefit.
It was said that the cods should be private. It kinda breaks the spirit of the program and makes the queue redundant since if you could use it more than once the codes become the new queue
Brilli-
.
Microsoft, Open-AI, Apple, Google and Meta are going hard to suck up everyone’s personal data to feed their AI’s … and you can bet that the governments of the world are going to see it all as well. No right to privacy if you use their products.
Just say NO.
Nillion, a startup building a decentralization platform without a blockchain, raised $20 million in seed funding.
Investment firm Distributed Global led the round, and Big Brain Holdings, Chapter One, GSR, HashKey, OP Crypto and SALT Fund were among the more than 150 investors. The company’s team includes founding team members from Uber, Indiegogo and Hedera Hashgraph, as well as leaders from Coinbase and Nike.
The project is based on a mathematical innovation invented by cryptography professor Miguel de Vega called Nil Message Compute.
“Nillion could open a new universe of use cases blockchain never imagined,” David Gan, founder and general partner of OP Crypto, said in a statement. “What proof-of-work did for blockchain, I think NMC may do for multi-party computation as a technology.”
While decentralized file storage isn’t new, many existing protocols such as Filecoin and IPFS use blockchains. Nillion will enable “the fragmentation and dispersion of data across the Nillion network of nodes which creates privacy while still allowing the underlying data to be computed on quickly.” Unlike blockchain technology, Nillion’s network nodes do not need to communicate with each other to store transaction data.
Seems very similar to
Big fan on Erik. I didn’t realise he was involved in Ai. He’d be an excellent person to get involved with Autonomi.
I post an interview of him a couple of weeks back where he talks about his AI venture:
What one of the tech channels on youtube has to say about AI
The “experts” lie - for professional profit.
This video makes me happy when I watch it.
MS 11 is the battlefield on which we will either succeed or fail, its where 80% of the users are who can benefit most from Autonomi.
Finding solutions around things like MS Recall client side scanning…, its only found on MS Copilot Plus installed on new MS Surface Pro machines (using RISC CPUs) and is only part of the challenge MS throws at you, key logging too is evil, both need to be turned off before installing Autonomi imo. Turn off Copilot too…,
Apparently the Intel IA64 CPUs do not have this MS Copilot Plus support capability…
I tried his venice.ai, at the moment their online access to the LLM delivers response results which is sourced from old, not updated with current facts data, so the response quality is less than good. They have work to do
I expect it takes quite a bit of funds to keep an LLM up to date, so this is an advantage to big tech and well funded companies. Will be the same for local open-source AI .
If AI models can be made super fast, then perhaps adding in our own news feeds as pre-prompt data, we can get around some of this limit.
Discussion between Glenn Greenwald and Tucker Carlson on free speech, censorship, hate, their relationships with others in their industry and similar topics.
4 posts were merged into an existing topic: Ukraine
The key is 1.58bit AI using Ternary +1 0 -1 matrix multiplication , and specialized learning sets focused on serving specific vertical market info needs AND Memory on Demand Services, which scale available memory matched to the AI workload specialized needs (varying in sizes to cover both execution and temporary caching memory needs).
Right now all AI Memory is siloed with fixed scale up sizes (ie Nvidia H/A100), with use it watch it sit their idle models, a really poor use of capital with shite ROI.
At the moment there are few/nada Marketplaces to serve workloads on demand from shared IT disaggregated Memory, CPU/GPU available resources. This challenge is both a market opportunity and very difficult to surmount using current SaaS models which feature lots of stranded resources (Amazon, MS and Google Hypescalers) front ended with hugely expensive memory caches.
For example these ‘hyperscalers’ excessively wasteful infrastructures are funded by Enterprises paying in part, on average, 50X the cost of Fast Storage media in the case of Amazon io1 or io2 Fast block storage services. Go figure…
Now that Big Company IT budgets are either frozen and shrinking, expect cost of infrastructure to be lower more rapidly via new nextgen IaaS innovations, given the ‘stock buy back’ fake economy of the FT500 EPS having nothing to do with actual Main St. Economy principles and everybody is now looking for cost savings, where energy bills monthly are typically 38-44% of IT infrastructure operating expenses.