MikeB

Blenderbot 3 in AI News

The bot is called BlenderBot 3 and can be accessed on the web. (Though, right now, it seems only residents in the US can do so.)

The bot is a prototype and built on Meta’s previous work with what are known as large language models or LLMS — powerful but flawed text-generation software of which OpenAI’s GPT-3 is the most widely known example. Like all LLMs, BlenderBot is initially trained on vast datasets of text, which it mines for statistical patterns in order to generate language.
...
However, these models also have serious flaws: they regurgitate biases in their training data and often invent answers to users’ questions (a big problem if they’re going to be useful as digital assistants).

This latter issue is something Meta specifically wants to test with BlenderBot. A big feature of the chatbot is that it’s capable of searching the internet in order to talk about specific topics. Even more importantly, users can then click on its responses to see where it got its information from. BlenderBot 3, in other words, can cite its sources.

8pla.net

Ben Goertzel with Desdemona the humanoid robot. AGI-22 Conference Invitation in Future of AI

https://www.youtube.com/watch?v=SQi0RtcpTwc

BitBoy Crypto interviews Ben Goertzel, and Desdemona the humanoid robot

https://www.youtube.com/watch?v=DYcHhToWVsU

WriterOfMinds

Quadruped robot (mechanical design and assembly) in Home Made Robots

I finally put some love into this old project this year and have some good pictures to show! Lots of build and design details on the blog if you want: https://writerofminds.blogspot.com/2022/08/ace-quadruped-30.html

The goal for this phase was to just get the robot to stand up, and I'm happy to say that I achieved that. I'm looking forward to working on it some more in the future and actually getting it moving.





Actuation is stepper motors (for now) via tendons.

MikeB

Nvidia's Omniverse ACE & AI Assistant in AI News

Nvidia's Omniverse ACE/Avatar Cloud Engine & AI Assistant demos for SIGGRAPH 2022.

Avatar Cloud Engine is comprised of four technologies processed on the cloud:
Animation AI, Conv AI, Vision AI, Recommend AI

Your characters face, data. Their filters and high-powered processing.

Technology timestamp: https://www.youtube.com/watch?v=Pev84SGO2r0&t=2685s

AI Assistant demo:
https://www.youtube.com/watch?v=Pev84SGO2r0&t=2726s

MikeB

Pattern based NLP in General Project Discussion

This is a project I've been working on for a few years. In 2019 I was testing out the theory on Pandora Bots, and this year I'm converting it to C/C++.

The main goal is to be small, fast, solid-state. Not algorithms, deducing/knowledge, learning...

It works by matching singular words, first through spell check, then into another full list of words which assigns each word a token (one physical character/byte symbol). There are approx 20-30 groups of symbols which all words match into. The 20-30 symbols are then used in pattern sentences. There are a few hundred pattern sentences. From each pattern sentence a broad intention can be gathered and this used in chatbot reponses.

The Chatbot responses are fixed to around 50-100 and are represented as a one character/byte symbol. These are used to create duplicate / cross language responses. They can also be voice recorded and assigned to each symbol.

It's aimed at time/space restricted Chatbots, for example in games, where processing needs to happen in milliseconds, but also take into account broad user input.

Size and Response times:
In 2019 in pandora bots (1000 words, 200 sentences) the size is ~500kb, and ~1 second response time.
In 2020 in C/C++ on an Arm Cortex M4 @ 120mhz it's ~159kb and 15-100 milliseconds.
In 2021 in C/C++ on a modern PC @ 2.6ghz with Binary Searching (set up time of 70ms), processing is ~1ms / sentence.

A few features:
 < 500kb including word databases.
 One sentence generally takes less than 1 ms to process.
 Limited chatbot responses make it easy to voice record and/or change personality.
 Private information stripped during word compression (no names/places).
 Fine differentiation of intentions, eg between: Wondering, Questions, and Directions - "can you speak english" "do you speak english" "speak english".
 Can count occurance of emotional words, logical words, burning-analyser words, light-sense words to reply in kind better.

For the problem of chatbots in games or time/space restricted platforms, it solves:
 Too much data or processing power required.
 Cannot change the personality/no personality.
 Cannot change the language/only one language.
 Chatbot escaping the topic due to bad intention reading.
 Chatbot returning bad views / knowledge calculation (only fixed responses allowed).
 Chatbot terrible voice synthesis (can pre-record all fixed responses including randomised duplicate recordings).
 Not being white box/solid state/predictable.

WriterOfMinds

Diagram of transformer algorithm in AI Programming

I ran across this. It claims to be a simple (and largely visual) explanation of a transformer architecture. I wondered if it might help out Locksuit or anybody else looking for a clear introduction.

https://github.com/markriedl/transformer-walkthrough

Freddy

What's everyone up to ? in General Chat

This topic was started as a general thread to share or muse about member's interests and projects.

LOCKSUIT

UK's Drone Superhighway in AI News

https://ibb.co/jbJ2Vkz

Where Drones Fly Free
Autonomous aircraft in the United Kingdom are getting their own superhighway.

What’s new: The UK government approved Project Skyway, a 165-mile system of interconnected drone-only flight routes. The airspace is scheduled to open by 2024.
How it works: The routes, each just over six miles wide, will connect six medium-sized English cities including Cambridge, Coventry, Oxford, and Rugby. They avoid forested or ecologically sensitive areas, as well as major cities like London and Birmingham.

A consortium of businesses will install a ground-based sensor network over the next two years to monitor air traffic along the Skyway. The sensors will supply information to help the drones navigate, removing the need for fliers to carry their own sensors.
The sensors will also feed an air-traffic management system from Altitude Angel, which will help the craft avoid midair collisions.
The UK government is considering future extensions to coastal urban areas like Southampton and Ipswich.
Behind the news: Project Skyway is the largest proposed designated drone flight zone, but it’s not the only one.

A European Union effort based in Ireland aims to develop an air-traffic control system for autonomous aircraft including those used for deliveries, emergency response, agriculture, and personal transportation.
In March 2021, authorities in Senegal granted approval for drone startup Volansi to fly its aircraft outside of operators’ line of sight.
The California city of Ontario established safe flight corridors for drones built by Airspace Link to fly between warehouses and logistics centers. The plan awaits approval by the United States Federal Aviation Administration.
Yes, but: Although Skyway includes a collision-avoidance system, it’s not designed to prevent accidents during takeoff and landing, when they’re most common. Moreover, it's not yet clear whether the plan includes designated takeoff and landing sites. “The problem is what happens when you're 10 feet away from people,” one aerospace engineer told the BBC.
Why it matters: Drones are restricted from flying in most places due to worries that they could interfere — or collide — with other aircraft. By giving them their own airspace, the UK is allowing drones to deliver on their potential without putting other aircraft at risk.
We’re thinking: Figuring out how to operate drones safely has proven one of the most difficult aspects of deploying them in commercial applications. This project is a big step toward ironing out the regulatory bugs and also provides a relatively safe space to address technical issues.

MagnusWootton

another robot build plan in Home Made Robots

So this robot is EM.  (dc motors)    And Im going to build it by putting down a couple of camping matresses and glueing them together for a bigger surface, then mark some precision dots with a black texta , then put in shish-kebab sticks and then form a polygon with super glued chux wipes in a big chain and then glueing to the matress for a polygon extrusion.   Im then going to put in the dc motors on little stands and put a plastic cavity box for the computer then the whole thing gets plaster poured and it should attach it all together.

I'm going to be using a GTX2070 desktop for the computer,  I'm going to get it out of the metal case and just have the motherboard in the plastic cavity for least weight takeup possible.

I've got an original motor driver circuit, I'm testing out and prototyping right now, but something wierd happened, and my voltage divider doesnt seem to be working properly, and that doesnt make sense!!!     Theres no way you can stuff a voltage divider up because they are so simple,   I'm getting current down the resistor alley and the short and it doesn't make sense,  only the short should be getting power!!!   I'll check it out again later.

So apart from that extreme wierd stuff up,     The build is going to cost me very little for the plaster,   I'll hopefully get it for less than $10 a kilo,  and maybe its possibly less than 10 kilos?     The dc motors Ive got a good deal for $5, and they have gears! so it'll be really strong. (These look really good, amazing bargain!!)
https://www.aliexpress.com/item/1005002647161186.html?_randl_currency=AUD&_randl_shipto=AU&src=google&src=google&albch=rmkt&acnt=576-373-4425&albcp=16259416291&albag=&slnk=&trgt=&plac=&crea=&netw=x&device=c&mtctp=&albbt=Google_7_rmkt&gclid=CjwKCAjwt7SWBhAnEiwAx8ZLauuati0Ifa4_3TJEl_Rx85ZK-_ddyhH0xgFGAhmVspgo6mrzezjy4BoC910QAvD_BwE&aff_fcid=3ba324d49ae147d5a6f12a3138c5fd33-1657640758244-05415-UneMJZVf&aff_fsk=UneMJZVf&aff_platform=aaf&sk=UneMJZVf&aff_trace_key=3ba324d49ae147d5a6f12a3138c5fd33-1657640758244-05415-UneMJZVf&terminal_id=fb0511f736444abfbda844636b6ea049&afSmartRedirect=y

And then you need couplers and ill be using allen keys to couple a perpendicular arm to them for the plaster to grip onto.  That and the angle position sensors, I'll be trying to make some home made hall sensors which are going to pick up the angle of the shaft against the magnets inside the actual motor, So I dont need to buy another magnet for it.


After the body is done,  just have to put the software on the gpu,  GPUs are the best for the computer vision,  so the computer vision is going to be really top notch,  I'm going to try and fit a physics engine on a single call to the gpu for maximum motor search tests,   I have a shader compiled and it just fit. (Just an early version,  But it hasnt got the extra realistic fixes yet,  which is going to be hard to fit ontop of whats there.)     First thing I'll do when I get the robot body done is get it to stand up,  the gears will help it.

The use case of the robot is going to be fully autonomous excavator, carry follower, search and gather, and guard and takedown and disarm.
Thats going to be fairly simple programming, but extremely useful and never before seen If I get it done.
Robots arent developed much,  So most ppl dont even know how useful they would be,  But I'm getting an idea about it now.

I'll be back with more If I get further in the build.

frankinstien

Interesting Machine Learning course in AI Programming

I took this Coursers course: Machine Learning, it uses Octave/Matlab which I never had used before. The course emphasizes the matrix and vector approaches where I'm now going to replace a ton of code with Matrix libraries.   Octave can use your GPU, but you have to compile it with the library to do that. But over all a good course that explains the math of machine learning clearly and how to implement solutions for large data sets. I would suggest taking this course before endeavoring on other literature, it really will help you understand what's being done on other fronts.