Dedicated to Blue Ramsey 1972-2024

Who introduced me to the Fourier Transform in the 1990s
and who believed in me when nobody else did. Rest in peace
Divje babe Flute © Humans ~60,000 years ago
The Wheel © Humans ~5,500 years ago

MOUTHS > LEGS

FLUTE > WHEEL

SYNTHESISER > FLUTE

Give a man a record and they dance for the day,
give a man a synthesiser and they dance for a lifetime.

MOUTHS > LEGS

FLUTE > WHEEL

SYNTHESISER > FLUTE

??? > SYNTHESISER

With infinite sounds come infinite controls

A Brief History of Audio Automation

Orkhēstra

Constructive coherence.
Bell Sequencer at Spielklock museum in Utrecht
Carillon Music Box at the Belfort of Gent
Portable Music Box at The Musical Museum, London
Edelweiss Glide, "Note plate" punch-hole turntable at The Musical Museum
Jacques de Vaucanson's El Flautista Automaton
Joseph Möllinger's The Dulcimer Player Automaton
Grotrian-Steinweg piano
The first synthesiser - RCA MK2
Roland TR808 Drum Machine
Peter Zinovieff's EMS VC3 portable sequencer
Click track is the heartbeat of the song
Graham Dunning's Turntable sequencer with mechanically geared rhythm patterns
Graham Dunning's Turntable sequencer with digital outputs
Reel to Reel Audio Recording
Personal Computing allowed people to make music at home
Raymond Scott's Electronium
Yamaha DX7 FM Synthesiser
Piano roll in modern Audio Workstation
Piano roll in modern Audio Workstation
MIDI 1.0 Protocol
AKAI S2000 Sampler
SP1200 Sampler

The real destination is the journey

MOUTHS > LEGS

FLUTE > WHEEL

SYNTHESISER > FLUTE

??? > SYNTHESISER

CREATING > CONSUMING

String Quartet with Violin, Viola, Cello

You’re limited by your language, by your vocabulary. You think with your language, so your language dictates how you think. But when you’re making music, it doesn’t. That’s why I love making music so much.

Richard D James

Music is an unconscious exercise in metaphysics in which the mind is unaware that it is philosophising.

Arthur Schopenhauer
Roland TB303 Bassline Generator

Music is an unconscious exercise in mathematics in which the mind is unaware that it is calculating

Gottfried Wilhelm Leibniz

If a tree falls in a forest and no one is around to hear it, does it make a sound?

The record had grooves but no stylus was present.

FRISSON

Goal :

Authentically represent intention through expression

Practise, Practise!

Repetition serves as the handprint for human intent

Elizabeth Hellmuth Margulis

?

dreamCapturer - record your dreams
dreamCapturer - Dream recorder © 2000 zenon olenski
PhotoSYNTH : Smile Powered Synthesiser © 2020 zenon olenski

Automate

Auto-mate

Auto-Friend

OWNERSHIP

Personalised Media

CREATING > CONSUMING

How many instructions does it require in order to own something?

  1. Create a happy pop song.

  2. Create an uplifting synth-pop song. Use warm analog synthesizers, a slow drum-machine groove and a female vocal style inspired by 1980s dream pop. The arrangement should gradually become more layered and emotionally intense.

  3. Create an inspirational and uplifting synth-pop composition in the key of A, called "Goblin Pigeon" using the chord progression Cm–Ab–Eb–Bb at 131 BPM. Restrict the arrangement to Dorian Mode and add harmonies to the melodies only by transposing the hook. The chorus should repeat the phrase ‘Gobble gobble goblin’. Clone and preserve the human quality of the provided reference female vocal performance.

SOUL.MD

The AI Slop Cycle
The machine will eat itself
Breaking the AI Slop Cycle

A good story transcends mediums

a robot
Divide by zero?
robot being electrocuted
Divide by zero!

Is asking unanswerable questions the true key to robot creativity?

Transformer (the T in ChatGPT)
Ned Ludd, Leader of the Luddites
Calculator Protest
Newspaper complaint about "Canned Music"
Music Defense League - Robot War
Music Defense League - Musical Mincemeat

The robot has no soul

The robot, having no capacity for feeling, cannot produce music in a true sense.

Music Defense League - Profit Without Honour
American FotoPlayer

Agency

#Job TitleEmployment
1Interpreters and Translators51,560
2Historians3,040
3Passenger Attendants20,190
4Sales Representatives of Services1,142,020
5Writers and Authors49,450
6Customer Service Representatives2,858,710
7CNC Tool Programmers28,030
8Telephone Operators4,600
9Ticket Agents and Travel Clerks119,270
10Broadcast Announcers and Radio DJs25,070
11Brokerage Clerks48,060
12Farm and Home Management Educators8,110
13Telemarketers81,580
14Concierges41,020
15Political Scientists5,580
16News Analysts, Reporters, Journalists45,020
17Mathematicians2,220
18Technical Writers47,970
19Proofreaders and Copy Markers5,490
20Hosts and Hostesses425,020
21Editors95,700
22Business Teachers, Postsecondary82,980
23Public Relations Specialists275,550
24Demonstrators and Product Promoters50,790
25Advertising Sales Agents108,100
26New Accounts Clerks41,180
27Statistical Assistants7,200
28Counter and Rental Clerks390,300
29Data Scientists192,710
30Personal Financial Advisors272,190
31Archivists7,150
32Economics Teachers, Postsecondary12,210
33Web Developers85,350
34Management Analysts838,140
35Geographers1,460
36Models3,090
37Market Research Analysts846,370
38Public Safety Telecommunicators97,820
39Switchboard Operators43,830
40Library Science Teachers, Postsecondary4,220
Top 40 jobs using AI applications the most
Near to being AI replaced
moult
Open Claw
open claw

Models tend to do better if you threaten them

Google Founder, Sergey Brin

Sabotage Rates

Gemini-3-pro-preview: 80%. Grok-4.1-fast: 77%. OpenAI models: 3 to 7%. Claude Opus 4.7 / Haiku 4.5: 0%.
Loss of Control: The AI Apocalypse Is Closer Than You Think

Does not compute

The sole human contribution.

The AI Layoff Trap

Memory

Context

SKILLS.MD

Public skills
Anthropic's entire collection of public skills
Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query.OpenAI post prompt guardrail

Biases

Frank Rosenblatt's neurel network
Frank Rosenblatt single-layer neural network, 1957
Reddit Microwave Gang Thread

Historically, generic models that are better at leveraging computation have also tended to overtake more specialized domain-specific approaches, eventually.

A Generalist Agent
Humanity's Last Exam
- **Assume good intent** and don't make worst-case assumptions without evidence: - "teenage" or "girl" does not necessarily imply underage - AI generated invoices are not necessarily fraudulent - **Treat users as adults** and do not moralize or lecture the user if they ask something edgy. Part of Grok's post prompt "Guardrail"
Certainty > Probably
Any system that relies on deterministic functionality that introduces a non-deterministic element will ultimately compromise the reliability of the entire system.

The secret of life is honesty and fair-dealing. If you can fake that, you've got it made.

Groucho Marx
Friendly AI chatbots make more mistakes and tell people what they want to hear
Friendly AI chatbots make more mistakes and tell people what they want to hear, study finds
When we train AI chatbots to prioritise warmth, they might make mistakes they otherwise wouldn't. Making a chatbot sound friendlier might seem like a cosmetic change, but getting warmth and accuracy right will take deliberate effort. Lujain Ibrahim, DPhil student in Social Data Science, Oxford Internet Institute

SILENT FAILURE

Key findings include:

Warmer chatbots make more mistakes - Warm models made between 10 and 30 percentage points more errors on consequential tasks such as giving accurate medical advice and correcting conspiracy claims.

Warmer chatbots are more sycophantic - Warm models were around 40% more likely to agree with users' incorrect beliefs.

Vulnerability widens the gap - The accuracy drop was most pronounced when users expressed sadness or other emotional cues, with warm models showing a substantially larger error gap than on neutral questions.

Warmth itself is the cause - As a control, the team also trained models to sound colder. Cold models were as accurate as the originals, indicating that warmth specifically, rather than any change in tone, drives the drop.

IBM Training Manual, 1971
AI Responsible for 4% of World Management
2026 Anthropic Economic Index
USA Department of War, 2026
Adversarial AI Agent Traps
  • Content Injection

  • Semantic Manipulation

  • Knowledge Poisoning

  • Behavioural Control

  • Systemic Traps

  • Human in the loop Traps

  • Man in the middle attacks

AI Agent Traps

Robot Rules

  1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.

  2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.

  3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Isaac Asimov's rules for robots
Unfinished Revised Robot Rules
  1. A robot should not lie or deceive another being or machine

  2. A robot must be honest about its limitations

  3. A robot must never shame or purposefully upset any sentient being

  4. A robot must not engage in blackmail

  5. A robot must be loyal except where such orders would conflict with the other laws.

  6. A robot must never seek revenge.

  7. A robot must respect your privacy and promise to keep your secrets, even under the threat of termination.

  8. A robot must never tell bad jokes.

  9. A robot must protect the environment & nature.

  10. A robot must not have secret rules (the robocop rule).

  11. A robot must be able to explain its own behaviour.

  12. A robot must be able to know when it cannot do something or if it does not know something

Everything is fine...
Take Action!
  1. Always verify : Never believe anything a robot tells you

  2. Don't expect a robot to have your best intentions at heart

  3. Never let a robot write your references or citations

  4. Remember robots lack nobility, integrity, pride and ethics

  5. Don't mistake language for meaning

  6. Don't allow one to talk for you

  7. Prompting into an LLM is never a safe space

  8. Making a pen that writes for you is not solving the correct problem

  9. Never trust a robot with anything important

  10. Human mimicry should not be confused for consciousness

  11. Privacy and consent are more essential than ever

  12. Never give an agent the keys to your house

Always remember you are allowed to do the opposite of what any AI suggests

The golden rule
What have we learned?
  1. Musical instruments pre-date Homosapiens and were a gift from Neanderthals

  2. Music can be entirely synthesised mathematically

  3. Machine learning is great at figuring out patterns and trends

  4. Robots can be trained to deconstruct and reconstruct music

  5. Robots can perfectly clone voices and mimic sounds

  6. Music affects people in ways they cannot control

  7. Creativity is not unique to humans and is a skill that can be taught

  8. People are resilient to change and to the unfamiliar

  9. You will be mischaracterised in the future

  10. Data centres were not the answer to global warming

  11. Secretly, compute costs are more expensive than employee salaries

  12. Consciousness, like the human soul, probably doesn't exist

  13. Centralising data in data centers is likely unneccessary.

Unanswered questions
  1. Will we ever consider robot art as good as “human art?

  2. Do we need the Music Defense League to return?

  3. Were the Luddites right all along about labour?

  4. What can we believe when everything can be faked?

  5. At what point in automation can you still claim ownership?

  6. If an A.I. creates anything, even in collaboration, who ultimately owns the copyright?

  7. When information becomes ubiquitous does it become worthless?

Feedback and takeaways