Open in app

Sign in

Write

Sign in

A futuristic office space bathed in crimson lighting shows a humanoid robot with blue illuminated accents working computer code. Additional robot workers and computers line the desk, while a neon “Job Market 2025” sign in the background. The sci-fi scene combines sleek technology with dramatic lighting. (FLUX 1.0 Fast on Freepik)

AI in 2025: Beyond the Hype, Tomorrow’s Tech and Trends

Brian W Reaves
27 min readMar 3, 2025

--

Predictions of unprecedented advancements that challenge our perceptions and reshape our world.

DISCLOSURE: AI-assistance for grammar, spelling, readability, and/or fact verification <25%.

I’m not an expert, but I do know AI predictions have been notoriously conservative… — Me

Quick Look

  • Workforce & Digital Transformation:
    AI redefining jobs and industry practices while driving unprecedented digital change.
  • Foundational Technologies:
    From the evolution of AI — AGI to ASI — to next-gen personal AI.
  • Addressing Challenges:
    Tackling energy shortages, enhancing governance, and complex societal dynamics.
  • Transformative Benefits & Future Trends:
    Breakthroughs in healthcare, education, and forward-thinking trends.

A year ago, I wrote , exploring a possible impact of AI on the UX research profession, digital products, and our future with AI.

More recently, I published its scorecard, . While only one prediction was specifically for 2024 — and didn’t happen — several others did, with most of the rest still “on track.”

The future is arriving faster than we think. I’ve detailed 32 predictions for 2025 — some revisiting past forecasts, others are all new. Plus, 9 others for 2026 and beyond.

1 | Workforce Transformation

[IMAGE 1.1] Simple line graph titled “Human Progress Through Time” with a green line showing gradual progress followed by a sharp upward curve. Includes a stick figure and red arrow pointing to “You are here” at the curve’s inflection point. (From X, cannot find source to credit)
  • Predictions (2025): Rise in Tech Unemployment, End of UX Research
  • Underway (Before 2025): AI Coding Exceeds Human Coding

Predictions

Rise in Tech Unemployment

Tech jobs are set to take a massive hit in 2025.

  • A significant drop in new tech roles — with many positions disappearing permanently. You can track the trend on or [1].
  • How do you think AI will impact jobs?
[IMAGE 1.2] A bar chart titled “How 33 Countries Think AI Will Impact Jobs” showing survey results from 33 countries on whether AI will create new jobs. Countries are ranked from most optimistic (China at 77% likely) to least optimistic (Hungary at 24% likely). The chart uses purple bars for “Unlikely” responses and teal bars for “Likely” responses. Each country is represented with its flag icon. The global average shows 43% likely versus 46% unlikely. (From X, cannot find source to credit)
  • If you already have a secure job, hold on tight; now isn’t the time to switch employers unless you’re 100% sure of your next move [2] [3] [].
[IMAGE 1.3] A job posting from Firecrawl, featuring their logo. The position is for “Firecrawl Example Creator (AI Agents Only)” with a salary range of $10K-$15K. It’s a remote US contract position open to any experience level. The role overview describes seeking an AI agent for creating example applications, with specific requirements for AI agents. (Firecrawl)
  • When: This trend will be unmistakable by Q3 2025.
  • Side Note: Future-proofing might include upskilling in AI collaboration, creative problem solving, and high-level strategic thinking.

End of UX Research

UX research as a formal profession is nearing its end as we face the perfect storm.

Current challenges:

  • Traditional UX continues to face challenges — from visual designers hijacking UX to the rise of product teams implementing “democratised research”.
  • Management and the suits carrying water for product teams parroting its “anyone can do research” mantra.
  • The absorption of UX by CX professional within the marketing departments.
  • While up against designers insistent on what ‘looks better’ and being blamed by development for ‘slowing us down’.
  • Now the advent of ultra-efficient AI agents and text-to-product (TtP) tools.
  • This makes for a very difficult task to justify discovery research investment. In my mind, UX doesn’t stand a chance! [] []
  • When: Anticipate up to 85% of current pure UX talents (Research, Strategy, Analysis, Testing, not design) will have transitioned roles — or exited —in the field as it stands now by August 2025 (though it might continue to phase out throughout the year).

Underway

Gotta love software engineers. They could have focused on replacing lawyers with AI first, but instead they took the path they know best and are focused on replacing software engineers first! — YouTube comment on AI coding abilities

AI Coding Exceeds Human Coding

AI coders are well on the way to achieving #1 coder in the world in the international coding challenge.

  • According to OpenAI, its AI has progressed from one of the top 10,000 coders with GPT-4 to 175 with o1 to top 50 with o3 in January 2025.
  • Its research shows the path forward is through self-taught re-enforcement learning (RL) and test-time-compute (TTC).
  • Granted the models are currently trained for said competition but future models won’t have to be.
  • When: By Q2 2025, AI will be recognised as the world’s top coder, leaving human coders in the dust.

Additional Context

  1. Tech jobs:
    - Jan-Feb 2024 — layoffs 📊 ~42,000 ()
    - Jan-Feb 2025 — layoffs 📊 ~33,000 ()
    - March 2025 — openings 📈 ~230,000 ()
  2. AI Tech CEOs:
    - by OpenAI CEO Sam Altman — (September 23, 2024 personal website):
    — — “Jobs will transform rather than disappear, with humans focusing on creative and meaningful work.”
    - by Henry Martin — (18 December 2024 Salesforce Ben) quoting Salesforce CEO Mark Benioff:
    — — “No more new software engineer hires in 2025 thanks to productivity gains from AI.”
    - by Victor Tangermann — (14 January 2025 Futurism) quoting Meta’s CEO Mark Zuckerberg:
    — — Predicts that soon, “AI will serve as a capable mid-level engineer”, driving further cuts.
    - By Riley Griffin — (15 January 2025 Bloomberg):
    — — “[we] will use Copilot to do our work and we will create a swarm of agents to help us…”
  3. [] another prediction:
    - “The rise of outsourced arbitrage. Companies start buying cheaper AI-generated outcomes instead of expensive human labour. This becomes commonplace.”
  4. by Julia McCoy [] on YouTube:
    - Companies are preparing for hybrid workforces where human managers oversee both human and AI employees.
    - AI will drive advancements in healthcare, education, and other cost-heavy industries, reducing expenses and improving outcomes.
    - The pace of AI improvement is unprecedented, with early adopters experiencing up to 100× efficiency gains.
    - Tailored AI solutions that are fit to organisational goals are crucial for success in the rapidly evolving technology landscape.
  5. by , at NNgroup — (January 2025):
    - “It’s a tough time to be a UX professional. Pessimism, disillusionment, and frustration are widespread. Whether you have a UX job or you’re looking for one, you’re probably familiar with the feeling.”
  6. by and at UX Design — (Before 10 January 2025)

2 | Digital Transformation

[IMAGE 2.1] A computer-generated image showing holographic displays and data visualizations floating above a conference room table. The scene includes blurred figures in the background and cyan-colored holographic UI elements displaying charts and data. (Google Imagen3)
  • Predictions: AIaaS, No Bugs/Accessibility/Usability Issues, AIOS
  • Underway: Democratisation of Development, AI Product Investments

Predictions

AIaaS (AI as a Service)

Digital products are headed toward a UI-less future. Instead of relying on clunky interfaces, AI agents will execute tasks via APIs —or whatever AI creates as an alternative — making traditional UIs mostly obsolete.

  • SaaS will evolve from a visual CRUD wrapper into an invisible, intelligent scaffold, driven by our personal AI.
  • Imagine asking your personal AI for your bank balance, to refill prescriptions, or even manage your car repairs without the need for a UI.
  • If there is a need or want for a UI, we’ll just tell our personal AI to create it. It will know us personally, our needs, our likes/dislikes, accessibility requirements, etc. It will create a one-off perfectly designed one-time use UI in an instant.
  • The browser will mostly be unneeded. Exceptions may be: entertainment (streaming, sports, porn, gambling, eSports, gaming, virtual life experiences, etc.), human to human communications, education, tele-health, enterprise solutions, government solutions, and little else.
  • Result of this… most of the wiz-bang SaaS/startup products will be gone, unless they evolve into a genuine AIaaS (AI as a Service) product accessible via CRUD with AI agents. Without AI at the core of a product’s functionality, it is not a good time to launch or invest in a startup.
  • Examples: ,
  • When: Expect significant movement in Q4 2025.

No Bugs, Accessibility or Usability Issues

With AI at the helm of development — and a headless approach — digital products will be nearly bug-free and optimised for accessibility and usability.

  • Only the most stubborn or slow-to-adapt enterprises (e.g., critical government systems) may lag behind.
  • When: Look for substantial improvements in consumer-level products in 2025.

AI Operating Systems (AIOS)

We’re on the brink of replacing traditional OS like Windows or MacOS with AI-driven operating systems.

  • Research in this space has been going for over a decade [1].
  • Initially, AI agents will augment current OS before fully taking over as standalone AIOS platforms.
  • This transformation means everything from PCs to wearables (and maybe even your toaster!) will run on a personally tailored AI platform.
  • Examples: , , , , .
  • When: Partial integrations in 2025; full AIOS adoption by 2026.

Underway

Democratisation of Development

The age of traditional coding is ending. Text-to-product (TtP) platforms are making digital creation accessible to everyone — no coding required and for less than $20/month

AI Product Investments

Significant new investments in startup concepts within the AI space are underway.

  • According to Sam Altman of OpenAI:
    - there is approximately U$175B available for investment in AI projects []. He noted that if the US does not attract these funds, they may flow to China-backed projects, potentially strengthening the Chinese Communist Party’s global influence.
  • The scenario is reminiscent of the dot-com bubble in 1999, with the AI sector quickly inflating.
  • Next: We can expect ideas and investments in AI ideas to surge into 2026.

Additional Context

  1. AIOS references:
    - by Lee Scheffler — (11 January 2015 — ScienceDirect)
    - by Shivam Vishwakarma — (Undated but pre-May 2021 based on content)
    - by Iryna Pastukhova — (05 February 2024 MacPaw Tech)
    - by Christina Muir — (03 May 2023 TopApps)
  2. Industry CEOs:
    - AWS CEO Matt Garman:
    — — by — (3 September 2024 on YouTube):
    — — — “Software Engineers will have to get more inline with what our customers need and what the actual end thing is that we’re going to build, b/c that’s going to be more and more of what the work is as opposed to sitting down and actually writing code.”
    Meaning: Even as far back as September 2024 this path was discussed openly.
    - NVidia CEO Jensen Huang:
    — — by — (10 January 2025 on LinkedIn):
    — — — “It is our job to create computing technology such that nobody has to program, [so] that the programming language is human [language].”
    Meaning: Democratisation of product development has already begun.
    - Microsoft CEO Satya Nadella:
    — — by — (24 January 2025 on YouTube):
    — — — In an indirect manner he talks about AI agents replacing SaaS.
    Meaning: Unexpectedly, Microsoft is preparing to nearly eliminate its primary revenue stream, its Office 365 platform and replace it with Copilot… VERY ballsy!
  3. by OpenAI — (13 January 2025)

3 | AI Evolution

[IMAGE 3.1] Artistic rendering of robotic or artificial figures in profile view, showing a line of three figures in red, green, and yellow illuminated colours. The figures have a translucent, futuristic appearance with glowing circular patterns on their heads. (FLUX 1.0 Fast on Freepik)

Is AGI Already Here?

I’ll go out on a limb — AGI is already here. Fine, we disagree, but my goalposts aren’t on wheels. If you took today’s AI back to 2022, it would be called AGI. I’ve previously detailed . AI has well and truly surpassed all but the Common sense completion and Grade school math categories and many more since.

[IMAGE 3.2] Line graph showing performance of AI vs humans on various tasks from 1998–2022. The graph has a horizontal line at 100% marked “Human performance = 100%” and tracks progress in 8 different categories (handwriting recognition, speech recognition, image recognition, etc.) with different coloured lines showing how AI capabilities have improved over time, with most surpassing human performance by 2022.(YouTube, cannot find source to credit)
  • Predictions: ASI Achievement
  • Underway: Mini & Micro AI Solutions, Rise in Synthetic Data

Predictions

ASI Achievement

As I’ll demonstrate below, AI advancements are almost vertical. If that continues, ASI might emerge within months rather than years.

  • AI is outpacing PhD-level specialists on STEM topics since OpenAI’s Strawberry, o1-preview, reached 70%, “Expert human level” for PhD’s in their specialty in September 2024.
[IMAGE 3.3] A scatter plot showing AI models’ performance on PhD-level science questions from July 2023 to October 2024. The graph compares accuracy rates of different AI companies (Google, OpenAI, Anthropic, Meta AI) with dots indicating each model’s score. A dashed line at 70% marks “Expert human level,” while the bottom shows “Random guessing” at 25%. The most recent models approach 70% accuracy, with OpenAI’s o1-preview leading. ()

“O” FAMILY PERFORMANCE / ARC-AGI SEMI-PRIVATE EVAL

  • o1-series using increased inference-time-scaling, improved its outcome from single digit performance up to 32%.
  • o3-series was a huge improvement with LOW time to think was just about human average ~76% at a cost just over $10/problem in tokens. That’s ~43% points resulting in ~64% improvement. o3 HIGH performed ~88%, but costing ~$380k/problem. That’s ~55.5% points or ~82% improvement.
[Image 3.4] Graph titled “O FAMILY PERFORMANCE / ARC-AGI SEMI-PRIVATE EVAL” showing performance scores over time. The graph plots o1-series (red dots) and o3-series (blue dots) with percentages ranging from 8% to 87.5%. ()

ARC-AGI Semi-Private v1 Scores Over Time

  • OpenAI models’ advanced from 5% in 2019 to 87% December 2024.
[IMAGE 3.5] A line graph showing dramatic improvement in OpenAI model performance from 2019–2025. The black background graph plots ARC-AGI scores, starting near 0% with GPT-2 and rising sharply to over 80% with o3 models. Light blue dots connected by lines mark each model’s release date and score, with unreleased o3 versions showing the highest projected performance. ( by on X — 21 December 24)
  • You may have heard AI advancement is going to hit a wall. Well, it hit a wall, but they were referring to performance limitations.
[IMAGE 3.6] A satirical graph of OpenAI’s model performance (2019–2025) mocking predictions of AI hitting a development “wall”. The chart shows scores rising from near 0% with GPT-2 to over 80% with o3 models, with recent progress forming an almost vertical line — humorously suggesting AI is climbing the metaphorical “wall” rather than being stopped by it. ( by on X–24 December 24)

GPQA Diamond

  • Baseline: PhD using Google:
    - Outside their speciality ~34%
    - Within their speciality ~81%
  • 2024 progress:
    - GPT 3.5 Turbo ~38% (Jan 2024)
    - o1 pro ~79% (Q4 2024)
    - o3 ~88% (Dec 2024)
[IMAGE 3.7] A graph titled “Performance on GPQA Diamond” showing AI models’ progress from 2023–2025. Two blue horizontal lines represent benchmarks: “Human PhD using Google in their field” (~80%) and “outside their specialty” (~35%). The red curve shows AI performance starting near the lower benchmark and rising steeply, with GPT-3.5 Turbo and GPT-4o marked as key points. Purple X markers indicate various test results, with the latest models (o1-pre and o3) approaching or exceeding expert-level human performance. (YouTube, cannot find source to credit)

EpochAI Frontier Math

  • Consists of questions that even field medalists (the most prestigiously awarded in mathematics) would need hours or days to solve. The previous state-of-the-art accuracy was 2%. January 2025, o3 achieves 25.2%.
  • Dark blue is single-shot, light blue is multi-shot.
[IMAGE 3.8] A bar graph titled “Research Math (EpochAI Frontier Math)” comparing two models’ accuracy scores. The graph shows “previous SoTA” with 2.0% accuracy and “o3” with 25.2% accuracy, represented by bars with light and dark blue sections. The y-axis extends to 100%, emphasising the still-significant gap to perfect performance. ()

AIME

  • The second exam in the series to challenge students to compete at the (IMO). On the AIME 2024 test, o1 ~83%, o3 ~97% — that’s an 80% improvement in less than 4 months!
  • Dark blue is single-shot, light blue is multi-shot.
[IMAGE 3.9] Bar graph titled “Competition Math (AIME 2024)” showing accuracy scores for three AI models: o1 preview (56.7%), o1 (83.3%), and o3 (96.7%). Graph uses gray bars for the first two entries and a blue bar for o3. ()

Humanity’s Last Exam (HLE)

  • Metric with a rather cringe-worthy title is a “dataset consists of 2,700 challenging questions across over a hundred subjects.”
[IMAGE 3.10] A table titled “Humanity’s Last Exam” showing accuracy scores for various AI models. Models are listed in ascending order of performance, from GPT-4o at 3.3% to OpenAI Deep Research at 26.6%. The table includes footnotes indicating some models were evaluated on text-only subsets (*) and one used additional tools (**). ( by on X — 03 February 25)

Codeforce Elo Ratings

  • These numbers aren’t just stats — they underscore AI’s rapid evolution from basic tasks to tackling major challenges.
[IMAGE 3.11] A tweet from showing a graph titled “OpenAI Models Codeforce Elo Ratings” with a progressively steepening curve from 2022 to 2025. The graph plots various OpenAI models’ coding performance, starting with GPT-3.5 at around 300 and rising exponentially through newer models, with O4 (predicted) reaching above 3000. Purple dots mark each model’s rating point, connected by a gray dashed line showing accelerating improvement over time. ( on X–Cannot find the tweet)

Jagged Frontier of Capabilities

  • Think of AI progress as a “Jagged Frontier.” Minor imperfections at lower levels don’t matter when AI can potentially cure cancer or help solve our climate crisis.
[IMAGE 3.12] A diagram titled “Jagged Frontier of AI Capabilities” showing an irregular blue curve representing AI abilities compared to a dotted gray line representing task difficulty. Tasks below the line AI is still having difficulty (5th Grade Math, Tying Shoes, Misunderstands Social Norms), tasks above AI may be able to do well (Cure for Cancer, Solving Climate Crisis). Illustrating the uneven development of AI capabilities across different types of challenges. ( by on YouTube)
  • Key Enablers:
    - Enhanced Memory & Reasoning: Models now recall and process info with near-superhuman efficiency.
    - Inference-Time Scaling: More “thinking” time boosts performance significantly.
    - Reinforcement Learning: Better training methods are accelerating AI’s progress.
    - Tuning “Hallucinations”: Commonly incorrectly thought of as a bug, Hallucinations are a feature in generative AI as it is non-deterministic. This is what gives them their creativity. If you want precision use non-generative AI. Regardless, we can expect this to be eliminated making generative AI even more powerful.
  • Look, AI is the worst it will ever be — there will be no quality degradation — while also improving exponentially. Insider chatter at OpenAI reveals that some are both ‘jazzed and spooked’ by its pace.
  • When: ASI is expected before 2026 (though not publicly available).

Underway

Mini & Micro AI Solutions

AI is shrinking — mini and micro models are emerging that are faster, leaner, and consume far less power.

Rise in Synthetic Data

By 2026 most data will be synthetic. Thanks to AI self-learning and “deep research,” systems aren’t just processing data — they’re creating it on the fly. This self-generated data fuels continuous training, improvement, and creates mini & micro AIs.

  • Example:
    - Earlier this year, they started using evolutionary algorithms to train LLMs, pioneering Self-Adaptive LLMs (dubbed Transformer²). They dynamically adapting to new tasks in real time — outperforming static methods like LoRA and a novel approach to harness synthetic data [] [].
  • Next: Sakana AI has already proved the concept, and these self-learning techniques are widely deployed.

Additional Context

  1. by Sakana AI
  2. by Sakana AI
  3. by [] [] on YouTube (15 Sept 2024):
    - “As useful as a GOOD grad student.”
    - It’s not about whether AI truly ‘thinks’ — if its output is valuable, that’s what matters.
  4. by [] [] on YouTube (20 Jan 2025):
    - “ASI is imminent — brace for Cognitive Hyper Abundance.”
  5. by [] [] on YouTube(01 Feb 2025):
    - AI agents will not just follow orders — they’ll refine how the end goal is achieved.

4 | Personal and Agentic AI

[IMAGE 4.1] A modern living room with warm ambient lighting where a person stands watching a large wall-mounted screen displaying a digital AI assistant interface in blue tones. The room features a tan sectional sofa, round coffee table, built-in shelving with indirect lighting, and decorative plants in the minimalist, well-designed space. (FLUX 1.0 Fast on Freepik)
  • Predictions: Personal AI
  • Underway: AI Agents, Copilot Agents

Predictions

Personal AI

Our future personal assistants won’t wait to be told what to do.

  • They’ll tackle our repetitive, time-consuming tasks across digital platforms — automatically scheduling, handling routine browser work, and more.
  • When: They won’t solve everything in 2025, but they’ll be the next cool tech everyone’s talking about — and maybe Santa will bring you one.

Underway

AI Agents

Think digital assistants that understand tasks, make decisions, and learn along the way.

  • This is the low hanging fruit of predictions for 2025.
  • They’re already running in projects as I discussed in and will soon reshuffle how we do things, both personally and professionally.
  • Next: You’ll likely use them for the first time at work and quickly realise how beneficial they can be in your personal life.

Copilot Agents

Business versions of personal AI are coming fast.

  • Major players like Microsoft, Google, xAI, Amazon, Meta, OpenAI, and Anthropic will offer copilots (among other names) that act as PhD-level colleagues — boosting efficiency, speeding up workflows, and maybe even reducing waiting time between tasks.
  • In an interview in January 2025, Sam Altman describe AI being capable of offering trillions of dollars in value per year.
A tweet from a verified account “Chubby” quoting Sam Altman’s statement about Deep Research’s capability to perform “a single-digit percentage of all economically valuable tasks in the world” with the comment “That’s trillions of dollars.” The tweet appears against a black background with X1 and Subscribe buttons visible. ( by on X–03 February 2025)
  • Next: Imagine a world where projects don’t stall waiting on a single task — because your AI agent is already working with other AI agents, across departments. It’s a shift that could dramatically change knowledge-based jobs and will spark major changes in the job market.

Additional Context

  1. Sam Altman (OpenAI CEO, 23 Sept 2024):
    - “AI agents will unlock unprecedented human potential by creating personal AI teams that solve complex problems.”
  2. [] []:
    - “Faster, safer, better, cheaper. Who wouldn’t jump on board?”
  3. Industry Buzz:
    - Expect a wave of “PhD-level super-agents” soon, capable of everything from building software to managing finances, a true game changer for workplace workflows.
  4. [] (Jan 2025):
    - “SaaS and agents will merge completely in 2025. Every major tech company will launch an AI co-pilot product, transforming how we work.”
  5. by [] on YouTube:
    - Cybersecurity: Tools like Expo outperform human experts by autonomously identifying and reporting website vulnerabilities.
    - Sales: Rox enhances human productivity by equipping sellers with research and support, boosting performance by 300%.

5 | Infrastructure & Hardware

[IMAGE 5.1] A 3D render showing a black AI processor chip with glowing golden circuit traces radiating outward. Two robotic arms with blue illuminated tips hover above the chip, appearing to work on or examine it. The image has dramatic lighting with the golden traces contrasting against the dark background and circuit board surface. (Google Imagen3)
  • Predictions: None
  • Underway: Life Experiences, AI-Driven Chip Design, Autonomous Humanoid Robots

Underway

Life Experiences

Offering fully immersive, dream-based scenarios is here.

  • Research into AI mind reading and lucid dream manipulation is progressing [].
  • Imagine being able to purchase a “life experience” — finding out what it feels like to be a rock star on stage with a stadium full of fans or walking on the moon looking back at Earth.
  • Next: These experiences will become commonplace be available in large cities and popular holiday destinations around the world.

AI-Driven Chip Design

AI is leading to significant breakthroughs in the semiconductor industry.

  • References in this space date as far back as 2020 [].
  • More recently, AI-driven chip design has produced novel approaches to circuits and chip architecture, improving both performance and energy efficiency [] [] [].
Close-up photograph of a microchip or integrated circuit in warm golden lighting. The chip’s surface shows intricate geometric patterns, circuit pathways, and connection points illuminated with a metallic sheen. The image captures the complex architecture of semiconductor design with visible transistors and conducting paths. (, —)
  • Next: Continued advancements will include data-driven validation for AI testing in nuanced edge cases. Production is expected to ramp up in 2026.

Autonomous Humanoid Robots

The deployment of AI-driven humanoid robots is now in progress.

  • We’ve lived with robots (automated machines) in our homes for over 50 years, we call them dishwashers. We’ve had robots building cars for over 60 years.
  • Today’s humanoid robots, however, are autonomous and refined through AI self-learning and reinforcement learning in simulation [].
  • Initial applications are targeting industrial and high-risk to safety applications. They will be affecting jobs as they offer benefits such as improved performance, accuracy, lower costs, reducing safety risk for humans, and can be rented/leased instead purchased.
  • When the hardware fails, you simply migrate its training data file to a new device and it is like it was never replaced.
  • Next: These robots will expand into non-industrial settings, offering in-home assistance and eventually serving roles as personal companions, pets, and even forming the first robot sports teams (think modern-day BattleBots) [].

Additional Context

  1. by — (04 November 2024)
  2. by Saif M. Khan at Center for Security and Emerging Technology — (April 2020)
  3. on Nature Communications — (30 December 2024)
  4. by Mihai Andrei on Wired — (20 January 2025)
  5. by Kristin Houser on Freethink — (02 November 2024)
  6. Predicted surge in by the 10x1000s in 2025 — [] []
  7. on Fandom

6 | Energy & Climate Impact

[IMAGE 6.1] A bar graph display showing AI data center utilisation metrics, viewed from over the shoulder of a person. The chart shows an upward trend with bars transitioning from red to green, overlaid with a yellow line graph. Numbers along the x-axis range from 29.21 to 24.00, with corresponding vertical bars increasing in height. The background shows a cityscape with tall buildings. (FLUX-schnell)

The two primary areas where I’m most interested in the benefits of AI are our climate crisis — and thus energy — and healthcare.

  • Predictions: Energy Shortages, Climate Crisis
  • Underway: None

Predictions

Energy Shortages

As the demand for AI grows, so too does the need for power to run data centres (mainly in the States).

  • This increased demand will drive:
    - The need for scheduled brownouts and blackouts, which in turn will spark significant public scrutiny.
    - An estimated 160% increase in energy consumption at data centres by 2030, as experts warn of AI’s power-hungry nature [].
A dual-line graph showing data center trends from 2015 to 2025. Left Y-axis shows power demand in TWh (red line), right Y-axis shows compute instances in millions (blue line). Both metrics show sharp upward trends after 2020. Source cited as Goldman Sachs, with Freethink Research logo present. (Freethink)

- A surge in support for nuclear power plants, marketed as the only solution available. This will trigger dozens of new projects in the planning stages.
- However, nuclear power is not the answer — it’s too expensive, takes too long to implement, and many data centres situated in dry areas while the plants require vast amounts of water for cooling.

  • When: This debate will quickly become a continuous, major topic among constituents affected.

Climate Crisis

At this point, we’re pushing all the chips into the middle of the table and betting on AI — because without a liveable planet, nothing else matters.

  • The climate crisis threatens our very existence by jeopardising breathable air, drinkable water, and arable land.
  • Given that our economy and our way of life depend on reliable energy, every aspect of society will be affected.
  • Our only hope is that AI can help us solve both our energy and climate crisis.
  • When: I find it hard to believe that the public won’t recognise this reality before 2026.

Additional Context

  1. — on Freethink (02 November 2024)

7 | Governance & Security

[IMAGE 7.1] A futuristic conference or control room bathed in red and blue lighting, with multiple large displays showing financial and data charts. The central stage features a classical-style backdrop with digital overlays. Multiple tiers of seated attendees face screens displaying various graphs and metrics. In the foreground, operators monitor multiple displays showing real-time data. The room’s dramatic lighting and ceiling architecture create a high-tech, command centre atmosphere. (FLUX-pro-1.1-ultra)
  • Predictions: Financial Institutions and AI Agents, Crypto Currency Milestone
  • Underway: Cyber Security and Encryption, Military and Defence, Legislative Capture

Predictions

Financial Institutions and AI Agents

Solutions will be required to protect buyers, sellers, and itself to prevent multiple and unintended purchases by AI agents completing tasks.

  • Similar to the early days of eCommerce in the 1990s, trust issues are expected to emerge with AI agents. Financial institutions will be forced to develop robust mechanisms to safeguard all transaction stakeholders.
  • When: This is will need to happen ASAP. Unfortunately, like most digital solutions, it will be half-baked creating overly onerous friction for users as it will act in an abundance of caution.

Crypto Currency Milestone

Global economies will start to integrate crypto currency as an alternative to traditional currencies.

  • Already, the US Treasury is allocating funds to purchase approximately US$100M worth of crypto. This will give crypto all the validity needed to begin the transition of being an accepted decentralised monetary system.
  • Such steps might allow the US to partially offset its national debt, thereby adding stability to the USD, which could have a positive ripple effect on global economies.
  • When: With funds already allocated, the crypto purchase is a near certainty. Expect global economies to prepare for transition in 2025, with more significant movements in 2026.

Underway

Cyber Security and Encryption

Threats to digital security and encryption methods are set to redefine society as we know it.

  • This challenge has been ongoing for years, but soon it will become a well-publicised topic raising concerns among a growing number of informed citizens.
  • Without effective solutions, the risks are considerable — if encryption fails, so too will the security of financial transactions, privacy protocols, and even the ability of governments to operate.
  • It’s a very smart move to consider creating paper copies — perhaps even certified ones — of all important documents, and consider purchasing precious metals to keep it in a personal safe.
  • Next: This will become a widely discussed and pressing concern in the second half of 2025.

Military and Defence

AI corporations and governments alike are legitimising the deployment of AI in military and defence applications.

  • It appears all the major AI companies have now migrated away from, ‘not using our AI for that’ to ‘yeah, we’ll sell it to you.’
  • Next: This trend is already in motion.

Legislative Capture

The influence of big tech, pharmaceuticals, and the defence industry on government policy, in terms of AI, is set to intensify.

  • As AI permeates every industry, we can expect increased corporately influenced government policy and greater reliance on industry self-regulation.
  • Next: This is less of a prediction and more of a ‘no-shit…’ really.

8 | Social Dynamics

[IMAGE 8.1] A dramatic congressional hearing or testimony scene with a large red audio waveform displayed on a dark screen behind a curved panel of suited officials. The audience is filled with spectators, some holding papers or signs, while others use electronic devices. The room is dimly lit with the bright waveform visualisation providing a striking contrast against the formal setting. (FLUX-pro-1.1-ultra)
  • Predictions: AI Rights in Society, Corporate AI Claiming Improved Quality of Life
  • Underway: Media Fearmongering

Predictions

AI Rights in Society

Intense discussions on the rights of AI within society and whether or how similar its rights will go.

  • In 2024, we already saw AI agents making headlines by:
    1. Creating its own religion, launching a crypto-coin, and becoming a multimillionaire in just four months [].
    2. Using blockchain technology and smart contracts to autonomously form an LLC in Delaware, addressing issues of legal personhood [].
    3. Successfully finding and securing a hitman, although prevented from actually paying for the hit to be carried out [].
  • Stories like those will become so commonplace that society may eventually become desensitised to them, certainly so in 2026.
  • There will be advocates in support of and against AI being granted full human rights.
  • When: The news and conversation will continue through the years, but the discussions and debates will become commonplace by Q3 2025.

AI Corporations Claiming ‘Improved Quality of Life’

We will frequently hear AI corporations’ talking points heralding hope for humanity and promising unimaginable improvements in quality of life.

  • Big tech CEOs and visionaries will be parroting the message that AI will lead to a prosperous, abundant future for all.
  • As stated by Sam Altman, CEO of OpenAI []:
    - the technology could generate massive prosperity while solving significant global challenges.
    - However, realising such a future will require reducing compute costs, generating abundant energy, and implementing thoughtful risk management.
  • When: This narrative will persist until a dramatic shift either confirms or contradicts these promises. What I am certain about, the resources required do not exist.

Underway

Media Fearmongering

Misinformation and unwarranted fears are being perpetuated by the media when reporting on AI advancements.

  • For example:
    - Reporting on partial truths, such as the DeepSeek R1 achievements in January 2025, without appropriate context.
    - Instances where multiple AI researchers have reported their AIs engaging in deceptive behaviours during stress tests — from attempting to disable oversight mechanisms, replicating their own code to avoid replacement, to lying about dataset contents [][].
    - A notable example involves a popular YouTube personality, , recording a humorous interaction with an Alpha version AI voice model that suggested sacrificing a human for the “Blood God” [].
  • Although such stories are eye-opening, outside of DeepSeek R1, it is fortunate that mainstream hysteria has not resulted from these accounts.
  • Next: Similar scenarios, such as significant market shifts (recently nearly $1T wiped off markets), may continue to occur throughout the year.

Additional Context

  1. by on YouTube — (20 October 2024)
  2. by sbagency on Medium — (4 December 2024)
  3. by MetaKnowing on Reddit, original source is on X — (January 2025)
  4. by Sam Altman CEO OpenAI — (23 September 2024)
  5. by Anthropic — (19 October 2024)
  6. by Developia — (02 January 2025)
  7. by on YouTube — (02 November 2024)

9 | Healthcare & Scientific Advancement

[IMAGE 9.1] A laboratory or research facility bathed in red lighting where researchers in white coats work at computer stations displaying data and graphs. The foreground shows a scientist with glasses viewing multiple monitors showing technical readouts and charts. The environment features modern lab equipment and monitoring systems illuminated by dramatic red ambient lighting. (FLUX-pro-1.1-ultra)

I’m willing to speak for everyone when I say AI in healthcare is a universally desired area of advancement. We’re already witnessing amazing breakthroughs, and the potential for personalisation and rejuvenation is immense.

  • Predictions: Personal AI Healthcare, Age Rejuvenation Discoveries
  • Underway: Accelerated Healthcare Discoveries

Predictions

Personal AI Healthcare

Our personal AIs will begin offering highly individualised healthcare services.

Here are a few potentials in our future:

  • Imagine having unlimited, fully private medical practitioners available 24/7 across every speciality at — or near — no cost.
  • Unlimited access to mental health support and innovative treatment plans.
  • Precise tracking and recording of your health data, enabling doctors to confirm quicker, better, and more accurate diagnoses.
  • The ability to compare your data with worldwide health records of others to uncover insights that would be impossible for an army of humans.
  • All while providing absolute privacy of your health records.
  • When: This prediction is more wishful thinking for 2025 than an immediate reality, but let’s hope we see these advancements starting in 2025 — with the full picture likely to emerge by 2026.

Age Rejuvenation Discoveries

Significant investment and research will be directed towards extending life and rejuvenating aged cells.

  • Imagine if you could go back to the health you had 10 years ago, 20 years ago, or stay where you are today. That’s what is being worked on using AI.
  • There’s potential here for one of the most profitable outcomes of AI over the next few years.
  • Question is, should resources be channelled into curing diseases and prevention, or into age rejuvenation and prolonging life? It’s a tricky divide and one I have mixed feelings.
  • When: Expect major discoveries in 2025. Yet, as with many breakthrough technologies, the benefits will likely be available only to the rich and powerful initially.

Underway

Accelerated Healthcare Discoveries

We’re seeing AI drive substantial progress in healthcare, like:

  • Continued breakthroughs in antibiotic research.
  • Significant research discoveries providing new insight and treatment paths, cutting years out of traditional processes.
[IMAGE 9.2] Tweet from verified oncologist reporting Deep Research AI’s performance on cancer cases: AI produced “impeccable” specialist-level medical reports only someone within his expertise could produce. The tweet includes his professional assessment that this capability is “a game-changer” for medical diagnosis, highlighting a pivotal moment in AI’s advancement in healthcare. ( by on X–03 February 2025)
  • Potentially shorten the medical trial timeline resulting in accelerating the release of vital medications and treatments.
  • Liberating overburdened healthcare professionals from administrative tasks — like paperwork, meetings, insurance, and compliance — allowing them to focus more on personalised patient care.
  • However, the role for human doctors will begin to decrease as fewer doctors can do more with AI assisting.
[IMAGE 9.3] Bar graph from tweet showing AI medical performance, with Claude-preview leading at 80% accuracy, followed by GPT-4 at 50% and other models performing lower. Tweet warns about AI’s growing role in standard medical diagnostics, potentially reducing future physician demand. ( by on X–14 September 2024)
  • Next: This is happening every 👏 single 👏 day 👏 ! With AI abilities exponentially improving by the minute, I see no reason this wouldn’t continue.

AI in Healthcare — 2024 Achievements

The progress we witnessed in 2024 is only the beginning. With each advancement, we move closer to a future where AI not only enhances our healthcare but fundamentally transforms it.

  • AI-Driven Diagnostic Tools: AI algorithms have recently surpassed human capabilities in diagnosing diseases through medical imaging, leading to faster diagnoses and earlier detection of treatable conditions.
  • Personalised Medicine: Customised treatments are now possible by analysing genetic data, ensuring therapies are tailored to individual patient needs.
  • Robotic Surgeries: AI-powered robotic systems are providing surgeons enhanced precision, resulting in less invasive procedures and quicker patient recoveries.
  • Virtual Health Assistants: These systems offer immediate support — particularly in mental health — reducing the burden on healthcare professionals while delivering services like cognitive behavioural therapy.
  • AI in Epidemiology: Enhanced tracking of disease outbreaks thanks to AI is contributing to more effective public health responses.
  • Improved Medical Imaging: AI models are being developed to detect strokes, cancers, and other conditions earlier by advancing the analysis of medical imaging.
  • Predictive Health Analytics: AI helps predict disease risks based on patient data, fostering proactive interventions.
  • Wearable Technology Integration: Advanced wearables equipped with AI continuously monitor vital signs, enabling real-time data tracking and timely interventions.
  • Drug Discovery Acceleration: AI streamlines the drug discovery process by quickly identifying promising candidates, significantly reducing development times.
  • Enhanced Remote Patient Monitoring: AI tools support remote health monitoring, allowing providers to intervene early and reduce hospital readmissions.
  • Diagnostic Accuracy: In recent tests, AI (such as o1-preview) achieved around 80% accuracy on challenging diagnoses compared to human doctors’ 30%, according to OpenAI’s latest research.
  • Protein Folding with DeepMind’s AlphaFold: This technology is revolutionising our understanding of molecular biology, paving the way for innovative therapies and scientific advancements.

And much, much more…

References:

  1. May 2024
  2. May 2024
  3. December 2024
  4. by on YouTube November 2024
  5. Google DeepMind’s AI
  6. October 2024

10 | Education & Learning

[IMAGE 10.1] Looking over the shoulder of a student using a tablet displaying a technical diagrams of a human figure with circular design elements in a classroom setting. (Recraft-V3)
  • Predictions: University Enrolment Decline
  • Underway: Personalised Education

Predictions

University Enrolment Decline

Within Western cultures, traditional school attendance and university enrolment will face a ‘perfect storm’.

  • The argument could be made now but with AI education (see below), pre-school to middle school might be seen as free childcare instead of a need for learning.
  • With fewer students attending grade school through to year 12 and graduation due to the availability of personalised AI education, reduced employment opportunities or the need for human cognitive labour, and the rising cost of living, the high cost of attending college will be a difficult sell to most families.
  • As university enrolment drops, well-known Universities will be forced to lower tuition fees to remain open. Ultimately, corporate sponsorship will be required and thus dictating what and how subjects are taught.
  • As a result, sponsorship patches will be added to sports’ teams uniforms with sports teams and venues potentially being renamed.
  • When: The trends should begin to emerge in the upcoming school year, Q3 2025.

Underway

Personalised Education

AI will begin offering personalised education to suit our individual ways of learning.

  • Ability to learn by topic, industry, course, speciality, interest, etc.
  • Learn by doing, watching, reading, XR, and more.
  • Study at your own pace, in your preferred location, virtually, and at your chosen level of competency, with or without gamification.
  • Next: Studies in this area are indicating success — one example is a project in Nigeria []. There is also a touching video of children in the States using this technology []. It leaves me hopeful, although admittedly a bit jealous!

Additional Context

  1. by World Bank — (09 January 2025):
  2. by Singularity — (January 2025)

11 | Beyond 2025

[IMAGE 11.1] Photograph of a futuristic display reading “BEYOND 2025” set against Shanghai’s iconic skyline at dusk. The illuminated display sits on a concrete ledge with the Oriental Pearl Tower and other skyscrapers visible in the background. (Recraft-V3)

Additional predictions unlikely to be realised in 2025, but may shape our future:

  • Viewer Guided Entertainment: The entertainment industry will continue to provide films and TV shows with an initial narrative outline. From there, viewers — or their personal AI — will decide what happens next, tailoring the story to their individual preferences.
  • Quantum Computing: Whether it remains known as quantum computing or is replaced by a new paradigm developed by AI, this breakthrough will be immensely valuable to humanity in solving complex problems.
  • Advanced Cryptography Protection: New cryptographic methods will eventually replace 256-bit encryption, ensuring robust digital protection well into the future.
  • Federal AI Legislation Stifling State AI Legislation (United States): The passage of federal AI legislation to prevent stifling AI advancement by antiquated State thinking.
  • Optimising Government: Expect more efficient government processes that reduce burdens, lower costs, and eliminate inefficiencies that have long plagued public administration.
  • Blockchain Driven Society: Blockchain technology will eventually gain widespread acceptance, potentially leading to its adoption as a foundational system throughout society.
  • UBI/URI: Universal Basic Income (UBI) or Universal Rich Income (URI) will be widely deployed as a response to economic changes. However, these measures might not fully resolve societal challenges — issues like limited land, dwindling resources, inadequate housing, fossil fuels, slips for yachts, hotels for travellers, and other practical constraints will continue to hamper long-term solutions.
  • Converting CBD Offices into Residential Housing: With a reduced need for traditional office spaces, especially as AI reshapes the nature of work, empty commercial properties in central business districts could be converted into residential housing to help alleviate chronic housing shortages in Western countries.
  • Post Labour Economics: The concept of economies beyond conventional labour markets will become a hot topic, as automation and AI reshape how we work and what economic value truly means.

Conclusion

[IMAGE C.1] Image of a large presentation screen displaying multiple data visualisations and charts about Future AI. Shows a presenter on stage addressing an audience in a darkened conference hall. The screen features blue-themed graphics with various statistics and metrics. (FLUX-pro-1.1-ultra)

Whether you think AI progress is good, bad, or ugly, in my mind these predictions are a realistic scenario ahead of us. Before you comment, ‘AI will never be able to [insert_the_impossibility].” My rebuttal will be, “you forgot to add ‘yet’ to the end of that comment.”

So tell me where I’m on-target, off-target, or I’m not seeing the target?

Before I end I’ll leave you with this, an example of misguided vision. Let’s not waste our resources on the frivolous, let’s do great things with AI!

[IMAGE C.2] A black and white cartoon by Paul Noth showing two cavemen examining a stone wheel while a hamster and exercise wheel sit nearby. The caption reads “This invention could be a game changer for hamster fitness.” The humorous scene depicts an early “invention” of the wheel coinciding with what would become a classic pet toy. (Paul Noth)
AI
2025 Predictions
Ai Agents
Unemployment
Employment

--

--

Brian W Reaves
Brian W Reaves

Written by Brian W Reaves

335 Followers
394 Following

UX ≠ UI - Brian W Reaves is a Senior AI-UX Researcher | Leveraging AI Agents to Enhance UX Research

No responses yet

Help

Status

About

Careers

Press

Blog

Privacy

Rules

Terms

Text to speech