Technology develops exponentially because the best available technology is used to develop its successor, repeatedly, so progress accelerates. Particular products develop on an S curve; it can take decades for a product to emerge and be superior out of 100s that fail, but if its prospects are good then investment rises, improvements made, costs fall and sales rise, then plateau when the product reaches saturation and/or better products emerge. But taken together, technology develops in a straightish logarithmic scale.
Everything is in patterns. It is a short step from pattern recognition to pattern prediction. AI systems are neural networks modelled on animal intelligence, collecting information with sensors, inputs, such as images of dogs and editing them in layers to produce an output of the word ‘dog’. Each layer has nodes. Every node collects information from every node beneath it, processed and passed on to every node above it.
But with AI every bit of information is digital, the process is mathematical, has unlimited space so can have trillions of nodes and trillions of layers, is millions of times more efficient, operates at the speed of light and can share perfectly with any other system instantly and form part of a greater system like the internet. Long before AI reaches the limits of physics it will be able to do anything a human can in seconds and trillions of times more besides. AI will do all cognitive tasks and control robots, and robots will do all physical tasks.
LLMs began the current wave of AI using neural networks to analyse vast databases and produce unique outputs as summaries, answers to questions and patterns according to preset criteria such as computer programs that calculate the most likely word, symbol, etc to follow the reference words that may number in the trillions. Most are adding an I/O system to use any modality; sound, light, etc. Development is faster in the digital world because manipulating bits is 1000x cheaper and easier than manipulating atoms.
Chat GPT1 was first launched in November 2022. There have been three more versions, each 100 times more powerful than its predecessor, a pace that will continue indefinitely. Chat GPT alone has increased in power by 1mX in less than two and a half years, and 100s of equally powerful competitors have been launched since then. LLMs are being succeeded by agents that will develop as fast.
AI is a universe technology, the most powerful technology ever and is in use with greater impact faster than anything previously. Nothing can stop the AI revolution because the economics are irresistible and we will have massive problems, economically, environmentally and otherwise.
The hourly cost of a British worker is £40. LLMs are £1 or £2 per month per person and studies show [2023-24] increase productivity by 30-40%. Robots can be as small as flies, as big as a container ship or open cast mine machine, designed for any form of physical work, in space, land or sea. Most humanoid robots are projected to cost $7k to $20k for a 3-5 year useful life which works out at 0.3-0.8p an hour. The cost to use a given level of AI falls about 10x every 12 months and lower prices lead to much greater usage.
LLMs and their derivatives dramatically improve productivity but are only as reliable as the reference base. When using an LLM with a large, reliable base such as STEM subjects or internal organisational information then it is at least as reliable as people. Problems arise with ideas and systems funded by advertising and monetised using clicks because dramatic, contrary and even untrue items get more clicks; they will need legislation. The need for valid and accurate databases in the public and educational spheres are becoming critical.
AI agents control subsystems such as LLMs, robots, etc. and are designed to perform tasks autonomously or do 1000s or millions of tasks [eg, write and send emails to all in a database] with a single, simple instruction. Currently, soft subject AI cannot be entirely accurate but, arguably, nor can people, so one chooses an outcome that suits them, but AI can have higher veracity. AI agent reference bases may be only verified research, can analyse it better and faster than any human then set objectives and make decisions independently, without constant human intervention. They can perceive and interact with their environment and retain a perfect record of all inputs, outputs and actions to improve and develop competence.
Robots are AI controlled in AVs, humanoid or otherwise to move, dig, carry, manipulate, etc. so they will do all sorts of logistics, farming, mining, manufacturing, warehousing, distribution and service work. Robots will work 10-18 hrs, day or night, stopping only for recharging and maintenance [with backup], then stored or moved to another task. Robots and agents will learn billions of times more efficiently than people and share learning with linked machines. But this is only the beginning. They will improve by multiples annually. Labour shortage will be a thing of the past because it will depend only on the rate of production.
When? 2025 is the year in which a critical mass of decision makers realise that this is not the old world with better technology but a whole new world transformed by technology. Ordinary people are not so aware of the big picture but are feeling the impact of job losses, automated workplaces, lower incomes and the underperformance of government monopolies. AI systems are not yet 100% reliable but usually much better than any human. AI agents are expected during 2025 and able to do almost any administrative task.
Tesla's robotaxi is planned to be in use in Austin Texas on June 1st 2025 and soon after in California and some other U.S states. The EU and China are fast tracking legislation to enable robotaxis to be in use during 2025 or 26. Tesla predicts that production in each factory will be one every five seconds, 24/7/365. AVs will replace private transport for commuting and occasional use reducing demand for parking, vehicles and public transport. AVs will make commuting leisure or work time adding 5 or 10% to national productivity when widely used, allow people to live further from their place of work easing pressure on housing and the transport infrastructure. No doubt AV buses, etc will soon follow. Utilisation per vehicle will be 8 or 9 times greater than personal transport, reducing the vehicle production, repair and insurance market by 80 to 90% and abolishing local parking. The automation of logistics will have a much greater impact in simplifying and reducing costs and allowing 24/7/365 transport anywhere, any time for anything. AVs will be cheaper to run and service because they will become commodities with massive economies of scale, no driver, few moving parts and AI agents and robots will be common in all industrial settings before everywhere else.
Six nascent transformative technologies; R&D, precision fermentation, health, power generation, Artificial Super Intelligence and robotics are set to transform our lives over the next five to ten years.
Precision fermented meat and milk proteins are available for animal and industrial use now. They use 5% of the land and costs. Sufficient investment is now in place to accelerate production and reduce costs significantly over the next few years. The technology is applicable to any biological product. Over 70% of human land use is for agriculture and this can virtually be eliminated because bioreactors can be placed within cities using no extra land.
More solar energy lands on the earth every day than needed to meet all human needs for a year. The technology to use this abundance is insufficiently developed to be cost effective everywhere as it requires storage and supply overnight. In some cold parts of the world, we would need two to three months of storage. But the growth in usage, development and investment is very high and we can expect the cost of energy to fall dramatically over the next five to ten years. Once it is more cost effective than alternatives, which it is in many parts, then it will replace alternatives and supply will become virtually limitless and free.
Health. A growing number, engaged in and observing the technological revolution, believe that health can be so improved that life expectancy will increase by at least a year for every year that passes [Longevity Escape Velocity] in the near future. Probably the most optimistic is Ray Kurzweil who believes this will happen in 2029. Kurzweil’s record of forecasting technological development has been at least 85% accurate between around 1980 to the present. His forecasts are based on his ‘law of accelerating returns’ and the progress of IT which is the key factor in technology. Several successful experts and forecasters predict LEV to be between 2030 and 2035; a few forecast around 2045.
The basic evidence that perpetual youthfulness is possible is valid and straightforward. Adults give birth to pristine offspring due to just four hormones present when the foetus is formless which can be replicated in the lab. Further study shows that modifying the number of hormones and the duration of exposure to them can cancel aging whilst retaining cell identity [Sinclair, 2016]. All living things have repair and regeneration systems that can be partially activated simply and cheaply. Some organisms are ageless. Turning these basic truths into a safe, reliable and cost effective therapy is of course more complicated, but there is no shortage of capital investment by 1000s of firms to do just that.
AI and R&D. AI is accelerating R&D and automation of research methods such as statistics and experiments. This is currently focused on STEM subjects because they have clear rules, an enormous body of online research and are mathematically based. The advent of AGI, some forecast 2025 but most with great confidence by 2027, will be revolutionary because it will be able to multiply its own development at least 10x annually and be replicated many millions fold. It will be as if any institution can have at its disposal an unlimited number of super smart researchers able to complete research in minutes not years, and not of samples that have to be replicated endlessly but of all cases published. The acceleration in technological progress will then be greater than any one person can comprehend.
Management. The revolution is already underway and causing redundancies on an unprecedented scale. Most can cope with current technology but next year it will be 100x more capable and 1000x more products will be available, a pace of development that will continue indefinitely. Currently in the UK 20% of working age adults are not working. 51% dependent on welfare. 70 to 80% of all local taxation goes to salaries and pensions. The national debt is unpayable. Spending exceeds income significantly. All this is obviously unsustainable.
But by 2026 we'll have AI agents and AVs causing more redundancy and unemployment. Whilst people struggle to adapt, humanoid robots will arrive causing even more unemployment. Most governments will resist modernisation to retain power, but the unsustainability of the current situation made much worse by deflation will force them to choose modernization and success or economic collapse, failure and oblivion; most will choose the former.
Conclusions
The state will be automated. It will be impossible to continue with the old ways. Change will be too fast for consensus around political ideologies and strategies plus the usual problems of knowing what people really want. Markets have always done this best. Democracy is far too inefficient for the AI era and open to abuse so it will be replaced by feedback mechanisms to balance supply and demand similar to those that will be developed for the private sector. The only viable government ideology then will be pragmatism.
Power will concentrate in an ever shrinking number of bosses of mega firms then be fully automated. Markets will continue because of the need for millions of products changing constantly although firms will rarely have employees. It is generally safer when power is in the hands of rival capitalists to offset each other and prioritise business. The drive to power will always exist but other than market forces blunting it, tyranny is pointless if the would-be subjects are robots.
People who don't work are mostly happy. We have a natural experiment in homemakers, the retired and workless and where they suffer it is usually due to illness, poverty and feelings of worthlessness. Illness and poverty will end. The sense of worth we get from work won't count when no one works and finds meaning in family, religion, hobbies, etc. although there will be exceptions.
Mass immigration from incompatible cultures has created subgroups within European and other nations. Unemployment and AVs means that people need not live near work or schools so will choose to live near family and like-minded people and form enclaves. Contact will be voluntary and conflict and criminality minimised because of a lack of competition over resources, outbreaks prevented by technology and robots and travel is easy, cheap and safe. National identity will decline not only due to mass migration but also due to the use of similar technology for the same ends with some cultural variations with no need for national governments. Organisations to develop children will be organised locally by parents themselves.
AI Supremacy
AI will take our jobs and we will be freed from sickness, poverty, violence and work. Basic safeguards should be that no machine be dangerous, wholly autonomous or designed to prioritize its own survival but mistakes will happen so powerful technology should be denied, where possible, to bad people and counter measures developed.
Many fear that AI that is millions of times more intelligent and powerful than us will go rogue and may regard us as obstacles. There is an element of anthropomorphism in this viewpoint. Biological intelligence evolved to survive individually, mostly by trial and error. We developed savage urges to hunt, aggressive urges to defend territory and access to mates. We can be ambitious for self actualisation, power and status. None of this applies to machines. Some will have avoidance systems, waterproofing, hard shells, mobility, etc [vehicles?] to prevent damage to itself, users and others but this is not defence, aggression or ambition; that requires a sense of identity and purpose. AI is designed to do work. They are mechanical systems controlled precisely by algorithms, unable to deviate from those algorithms which may incorporate trial and error but not necessarily or at the expense of people unless it is a weapon. That some actions are unexpected and seemingly original is due to our limited understanding not unlike a shelf we’ve just put up falling down.
Some AI systems currently emulate human-like desire, intention, etc because they are trained on human material. Mobile systems will clean, carry and interact and some to counsel, advise and inform. Only some that work in, for example, factories and hotels will need to work as a team but that will be centrally controlled. The most powerful systems to, for example, control billions of robots, conduct huge experiments or manage factories will be immobile, with specific I/O ports and language and rely on separate systems including power that can be turned off. It is foolish already to have huge dangerous machines that cannot be stopped in an emergency.
Many cleverer and better-informed people than me are far more concerned about the dangers of AI so I will be very interested to read comments from people like this.