Forem: Infinite Waves

Game AI series: 'Seed' by Klang Games - Part 1

Infinite Waves — Fri, 16 Apr 2021 09:32:00 +0000

The year is 2142. You have just woken up after being cryogenically frozen for over 100 years. You're in Berlin. Who are you? Or, rather, who were you? You were a game developer, remember? The brain fog is clearing and you really want to brush your teeth, your mouth is dry and your breath is stale. You swallow that tiny bit of century old pizza that was stuck in your teeth.

You seem to be in some metropolitan offices, you see this logo everywhere:

https://www.klang-games.com/seed

You look outside the window and nature seems to have reclaimed much of the city. Vines and tree roots have grown and cracked through buildings. An industrial-green symbiotic hybrid of architectural structures. Elk run wild through the low-hanging mist. You feel simultaneously at home and completely lost, abandoned, but liberated.

And here, we break the fourth wall to orientate you. You must still be pretty sleepy after all that time sleeping in the fridge. That midnight snack turned into a very long nap, didn't it?

Inspired by how well Klang Games have set the scene with backstory and lore for *'SEED'*, their (in-development) epic-looking Massively Multiplayer Online (MMO) simulation game, I thought I would emulate their proposition of leaving Earth to go to Avesta, the planet scheduled for colonisation.

"Seed is a virtual world that allows players to self-govern and participate in a real player-driven economy in a vast, sandbox environment. Seed also examines the idea of what it is to be human, focusing on the philosophical concepts of futurism. A deep experience, breaking down the constraints of what it means to play a game."

Klang Games

Disclaimer:

SEED is still under development and may be subject to change.

We'll use the narrative of this attempted journey to the new utopia as a way to explain the game and discuss the AI and what goes on under the hood.

I spoke to their excellent game AI engineer Alessia Nigretti to get some insights into the the team's development process, the technology unpinning it, and of course the game itself, SEED.

The quotes prefixed with "Alessia" below are actual quotes transcribed from the interview I had with Alessia Nigretti about SEED.

This will be a multi-part series. Subscribe (for free) to get access to all parts, delivered to your inbox upon release:

Subscribe now

I want you to imagine that it's the end of days, Earth is over, finito (if you're of a Doomsdayer disposition you might not have to have to stretch the imagination too far).

To be clear this meta-narrative is of my own invention and should be treated as a fun spinoff separate to the world and story of Klang's 'SEED', but obviously is directly inspired by it.

If you're on-board so far please share your mission to recruit others:

Back to the story...

The walls of the office are full of plans. What was the plan? You remember you were part of something.

Seed Dev Blog: Vol 1

The AI system boots up on your computational wrist strap. The disembodied voice introduces itself as your AI accomplice, Alessia Nigretti:

Alessia:

"I studied Computer Science and Artificial Intelligence in Brighton, UK, for three years. During that time, I worked a yearlong internship at Unity as a Technical Evangelist.\ Given my background and my interest in AI, I started working with Unity's Machine Learning team on ML-Agents, in addition to other non-AI features.\ While I was in my last year of university, I was looking to relocate, and looking for a project that interested me. Klang really sounded like it was hitting exactly what I was interested in: making an entire game revolve around AI, using it as a way to describe real emotions, and as the hook to get people attached to the game, to get people really involved in the game."

Your wrist just started talking. At first, it spooks the heeby jeebies out of you. You sense your heart rate spiking. You settle down and ask:

"Are you... real?"

There is no response, and besides, you're not even sure if you are real.

The last you remember the whole world was turning hyperreal, a blurred physical-digital simulacrum constructed primarily of cat memes and infinite TikTok loops, and you wonder if Bitcoin ever did take over the financial infrastructure of the world?

On all of the screens in the office a video plays endlessly:

Then it makes sense. You know who you are. You are more than a game developer. You are a SEED pioneer. The game you were developing became retrospectively prophetic, it became a reality. You are flooded with a concoction of neurotransmitters that, unknown to you, was instigated by a Computer-Brain Interface embedded in your skull.

"Hey Alessia, what's happened to Earth?"

Alessia:

"The story for now, is that human beings have basically destroyed the planet as we know it, and were forced to move onto a new planet: Avesta. It has similar conditions to Earth, but we have to nurture it to get it to a state that is liveable."

"Just like the game." You whisper to yourself.

Tingling with waves of motivation-provoking dopamine you hurry around Klang Games HQ and gather an inventory. You grab a satchel to pack water canisters and enough candy to satisfy a small elephant. You slide out of your hibernation slippers and lace up a mean pair of kevlar boots. Evacuation, if you're not too late, is your primary goal.

Source: SEED

Encouraged by the dream of Avesta, the new world, you ask:

"Alessia, what is possible on Avesta?"

Alessia:

"You can of course go rank up in the society pyramid, but that's only one choice. As a player, you can decide to go very hardcore, if that's what you want your characters to do in the game. But alternatively, you can also choose to be one of the people focused on nurturing your family, your Seedlings, your relationships."

Source: SEED

Your eyes are still blurred and everything looks slightly pixelated.

You access a terminal on one of the computers and begin looking for departure locations. Within a couple of minutes you access a launch schedule. There is one rocket left, 6 miles away. It leaves in 42 minutes.

Source: SEED

You're still euphoric from the neurochemicals blasting your synapses, which you now interpret as severe panic as you consider the possibility you might be left as the last person on Earth.

Source: SEED

To keep updated with this series make sure to subscribe to this newsletter to carry on your journey next week:

Subscribe now

Until the next part in the series here are some great resources to check out:

The Life Cycle Podcast

https://www.thelifecyclepodcast.com/

Theory - SEED Blog

https://www.seed-project.io/theory

SEED YouTube channel:

https://www.youtube.com/c/Seedmmo

SEED Dev Blog:

https://www.seed-project.io/dev-blog

SEED Discord:

https://discord.gg/XJnKyUaY

Become a pioneer with email alerts:

https://www.seed-pioneers.io/become-a-pioneer

Sources:

Klang Games

WIP Thursday Vol 1 - Seed Dev blog

SEED Pioneers: The Next Step for Humanity

AI Snapshot: PIZZANATOR; Intergalatic Sea Shanties? DARPA Dogfighting #11

Infinite Waves — Mon, 29 Mar 2021 22:14:47 +0000

Issue #11 of the Infinite Waves Newsletter

AI snapshot is the Monday issue of the Infinite Waves newsletter. AI news, developments, and points of interest from the previous week.

Feature:

DARPA Artificial Intelligence dogfighting

Are we not already good enough at killing things and blowing stuff up, really? We need AI to help us kill each other better? But let's face it, the military are already deep into AI and there is exactly zero chance they'll decelerate their efforts in employing it any time soon.

Doggyfighting DARPA 🐶 🔫 🛩

The US Defense Advanced Research Projects Agency (DARPA) has a programme it likes to call Air Combat Evolution (ACE, obviously). As part of this programme they have been running simulations that pit AI-controlled fighters against each other in aerial combat know as 'dogfighting'.

They ran a very unfair 2 vs 1 simulation, two 'friendly' Lockheed Martin F-16s versus a single 'enemy' F-16 (poor AI needs some damn backup).

What's the ultimate aim?

DARPA expects to see unmanned air vehicles (UAVs) taking care of dogfighting autonomously, leaving human pilots to concentrate on "higher-cognitive battle manager decisions" (So... they'll be playing COD?). DARPA is attempting to show that Artificial Intelligence-controlled UAVs can be trusted in the sky battles.

What's Lieutenant (Colonel) Dan got to say about it?

"These new engagements represent an important step in building trust in the algorithms since they allow us to assess how the [artificially intelligent] agents handle clear avenue of fire restrictions set up to prevent fratricide,"

Colonel Dan Javorsek - Programme manager in DARPA's Strategic Technology Office

"This is exceedingly important when operating with offensive weapons in a dynamic and confusing environment that includes a manned fighter, and also affords the opportunity to increase the complexity and teaming associated with manoeuvring two aircraft in relation to an adversary."

Sorry Tom, automation is coming for your job too

Does this mean there is going to be an all-Artifical Intelligence Top Gun 3 cast, without Tom Cruise? Without-Tom-Cruise? I don't think the one-liners will sound half as good in binary, but that's just me.

News 📰

Did robot pizza delivery drivers just become sentient? 🍕 No, not quite. The Pizzanator isn't gonna show up at your door any time soon (cheesy joke, literally). But Dominos Pizza has taken a slice from the AI world by using services provided by the AI company Datatron (sounds (d)ominous 😳).

It will allow Domino's to improve functionality and optimisation with its AI models, with the main aims being to improve in-store operations, find untapped revenue opportunities, and enhance the customer experience.

"Machine learning models can provide significant value to an organization in several business applications, but without a solid ML operations pipeline, that value cannot be truly realized."

Zack Fragoso - Data science and AI at Domino's (Manager)

This story came from a site called Pizza Marketplace. Who'd have known such a thing existed? The "premier online destination for C-level pizza executive leadership seeking cutting-edge intelligence for their multiunit restaurant concepts", according to their site, fascinating. Anybody else feeling hungry now?

Will NVIDIA be shaking in their graphically processed boots? Probably not, but Flex Logix will be looking to steal some marketshare with a cool $55m in Series D funding. The battleground is "inference at the edge", where a trained neural network is used on a device to make actual predictions. Geoff Tate, Chief Executive, said "the company can be successful both licensing its intellectual property to other chip designers, as it already does, or via sales of its forthcoming inference chip, the InferX X1."
Choo Choo, railway company's savings are off the rails. 🚂 Network Rail have made cost savings of ~£30m , thanks to the use of machine learning. Not too shabby. The company utilised the construction risk forecasting platform nPlan to help it make the savings.
Bringing history to life. Nathan Shipley, a San Francisco-based artist, has produced some pretty freaky (but kinda cool) images of historical figures using AI, to show what they might have looked like in the flesh. The figures in question lived before the advent of photography, but I'm sure Henry the VIII would have loved the Polaroid camera, because it's so easy to cut off heads.

Ludwig van Beethoven (I think he'd be pleased with that)

You know who she is - NGL I thought she'd be hotter

Bill (Bailey) Shakespeare

Photo credits: Nathan Shipley

See the full list on My Modern Met

Tweet of the week 🐦

A bit of curve ball from my boy Lex Fridman, the MIT AI researcher and podcaster:

Lex Fridman

@lexfridman

One day a human child born on a distant planet will have to be taught about a place called Earth from where humankind originated.

19:29 PM - 26 Mar 2021

How do you explain to that alien kid what a Sea Shanty is? They don't even know what the sea is. Although, they'll probably have VR headsets so maybe they'll be playing pirate simulator and know all about the seven seas. 🌊

Something free 🙌

Event website

DataRobot's AI Experience Worldwide is a free virtual conference for those seeking AI success. We're bringing together business leaders and data scientists to share insights on building an agile AI-driven enterprise.

Sources:

The Worldwide Artificial Intelligence in Military Industry is Expected to Reach $11.6 Billion by 2025

Domino's, Datatron partnership amps up AI, machine learning forecasting

Flex Logix has two paths to making a lot of money challenging Nvidia in AI

Artist Uses AI To Recreate How Famous Historical Figures Would Look Like in Real Life

How Network Rail is saving millions by using machine learning

The Hunt For Transformational Growth event

Stop switching tutorials

Infinite Waves — Fri, 12 Mar 2021 12:00:32 +0000

As I'm learning Python this time around it is with the purpose of being able to understand and complete some AI projects.

This is the first thing that helps. Have a clear goal/endpoint in mind. What is it you want to be able to do or build?

I have tried a bunch of resources over the years for learning Python. I never got too far because I didn't have a clear goal, and I was more focused on web development.

I am now using Kaggle, for reasons explained in the above link. It is specifically geared towards Data Science/Machine Learning, so it makes a lot of sense to do their courses, as it segues right into the sort of projects I want to be doing.

Just now I have been doing the Python mini-course, and at a certain point the tutorial didn't make sense to me. The first instinct is to assume:

"Well, this tutorial is obviously BS. What idiot wrote this? It's clearly not simple and obvious enough."

"Well, I'm obviously super dumb I need some baby-style Python-by-numbers tutorial."

But I carried on through the tutorial and came back to that part and went over it again. Instantly it was all pretty clear and made sense the second time round.

It made it easier because I have definitely decided Kaggle is the way I'm going. So I kept confidence it was the right thing, and looking for something else wasn't a consideration.

Perhaps at these points we'd be tempted to assume this tutorial 'just isn't right for me', or 'I bet I can find one that explains things better.'

But, in the end, as long as you get a course that is well regarded it'll probably be fine. There will of course be variance in clarity and styles. But there will be trade-offs in all tutorials, documentation, and educational courses.

The moral of the story is pick your resource and stick with it (that's not to say don't do a bit of vetting and trialling early on).

In most cases you'll learn more struggling when you don't understand part of the tutorial than jumping ship to the next one.

AI Fundamentals Series: Algorithms

Infinite Waves — Thu, 11 Mar 2021 13:56:52 +0000

Continuing on from the first instalment in the AI Fundamentals Series I want to provide some basic definitions of terms I have, and will continue, to use. In the spirit of trying to assume no prior knowledge, the aim is to keep everything as simple as possible to pave the way for more in-depth understanding at a later point.

Algorithm

In computer science a broad definition is:

"Any well-defined computational procedure that takes some value, or set of values, as input and produces some value, or set of values, as output. An algorithm is thus a sequence of computational steps that transform the input into the output.

We can also view an algorithm as a tool for solving a well-specified computational problem. The statement of the problem specifies in general terms the desired input/output relationship. The algorithm describes a specific computational procedure for achieving that input/output relationship."

(Introduction to Algorithms, Second Edition)

In this way we can think of an algorithm, metaphorically, as a list of directions that takes you from your start point to your destination if you had to go on a journey. A list of steps. But it also defines the relationship between the input and the output. A set of instructions, to put it another way. A recipe is often given as a common example of an algorithm.

Artificial Intelligence Algorithms

Artificial Intelligence is the broadest definition of what we are trying to do when augmenting human intelligence via computers. Under this umbrella there are subsets. Machine Learning (ML) is a subset of AI, and Deep Learning is a subset of Machine Learning. Although AI and ML are often used interchangebly.

Machine Learning:

"The study of computer algorithms that improve automatically through experience and by the use of data."

Deep Learning:

"Is representation learning: the automated formation of useful representations from data."

All Deep Learning is Machine Learning, and all Machine Learning is AI.

There are different techniques and algorithms at the different layers, and this categorization is not exhaustive. In the next issue of this AI Fundamentals series these topics, definitions, and their utilities will be broken down further.

The progression of AI algorithms

The AI algorithms that are used today are more sophisticated than was used in the 1970s and 1980s. The algorithms were mostly hand-coded in what is referred to as Good Old Fashioned AI (GOFAI), as Nick Bostrum points out in his book 'Superintelligence'. Whilst effective in very narrow domains (playing chess for example) they struggled to map onto other problems.

They were also fragile and would not return expected results if the programmers made even slight errors in their assumptions (P.9, Bostrum).

"One reason for this is the 'combinatorial explosion' of possibilities that must be explored by methods that rely on something like exhaustive search."

(P. 7, Bostrum)

Failures occurred when complexity increased beyond simple instances. Now, with increased compute power, we are able to implement algorithms that, effectively, write themselves. The initial algorithm is programmed but then it updates istself as it learns from data.

In some sense the first wave of AI, classical AI, worked by identifying a problem, a possible solution, and implementing the solution, a program that plays chess [for example]. Right now we are in the second wave of AI. So instead of writing the algorithm that implements the solution, we write an algorithm that automatically searches for an algorithm that implements the solution. So the learning system in some sense is an algorithm that iself discovers the algorithm that solves the problem.

(Joscha Bacha)

Today some Machine Learning algorithms behave more like the brain in the form of 'Artificial Neural Networks' (ANNs), artificial 'neurons', links in a chain, that can be more finely tuned or produce degrees of inaccuracy and still function, rather than a full-system crash.

This is, after all, one of the aims of AI. To reproduce, synthetically, what the brain can do, what we can do.

If you're not already signed up subscribe by clicking the button below to get the next issue in the AI Fundamentals series:

Subscribe now

Sources:

Introduction to Algorithms, Second Edition. Thomas H. Cormen Charles E. Leiserson Ronald L. Rives.

MIT Deep Learning Basics: Introduction and Overview with TensorFlow

Machine Learning Wikipedia

Superintelligence: Paths, Dangers, Strategies

Joscha Bach: Artificial Consciousness and the Nature of Reality | Lex Fridman Podcast

AI Fundamentals Series: a basic Artificial Intelligence definition

Infinite Waves — Thu, 04 Mar 2021 14:59:56 +0000

What is Artificial Intelligence? 🤖

Below is a broad, high-level definition that actually breaks down into two definitions (it’s definitions all the way down 🐢).

The definition was constructed by Jeremy Kepner and Vijay Gadepally, course instructors at MIT (source below).

Narrow AI: The theory and development of computer systems >that perform tasks that augment for human intelligence such >as perceiving, classifying, learning, abstracting, reasoning, >and/or acting.

General AI: Covers the scope of narrow AI but is fully >autonomous.

Today this is possible because of the convergence of High Performance Computing, Big Data, and Computer Algorithms.

AI ‘Series’ posts

Within this ecosystem (Infinite Waves) of Artificial Intelligence discovery and comprehension there will be off-shoots to explain some of the conceptual underpinnings of AI. Exploring different branches of AI in an attempt to produce the fullest possible picture of the Artificial Intelligence tree. History, companies, people, algorithms, new developments, and likely roots the field will take.

These off-shoots will be called series. The present series, AI Fundamentals, is about digging to the essence of AI, conceptually. Providing explanations, definitions, and general theoretical understanding.

The start here is but a simple seed, with the intention of growing increasingly more detailed and comprehensive in subsequent posts.

Sources:

MIT RES.LL-005 Mathematics of Big Data and Machine Learning, IAP 2020

Instructors: Jeremy Kepner, Vijay Gadepally

Reproduced from my AI newsletter post:

AI Fundamentals Series: a basic Artificial Intelligence definition #2