Brian Wang has had over 90% accuracy with public predictions on public websites like Metaculus and Manifold.
OpenAI O3 model has shown that greater than human PHD reasoning is possible but with about 30,000 GPUs taking 30 minutes to infer the answer. I forecast an increase to 1 million GPU AI training from 100,000 at the end of 2024. There is continuing improvement in the efficiency and effectiveness of AI training. Advancements will improve AI beyond human experts and lower the costs for inference to less than $100 of compute instead of $10,000.
Google VLOGGER and Microsoft Vasa-1 can make convincing animations of faces and voices from a still photograph and an audio clip.
AI video and AI coding will continue to make leaps in capability that will be highly valuable.
In 2025, the main predictions are around how much AI is improved and deployed.
Will there be an awakening of vehicles as robotaxi beyond 200,000 vehicles ? Currently, there are some robotaxi that do not have in vehicle supervision. There are about 2000 such vehicles around the world. There are Waymo, Apollo GO and other companies using Lidar and other sensors.
X will have X Payments (Paypal like financial services) in 2025. This should happen in the first half of 2025.
There will be X TV and expanded and much improved X Video.
Over 150 launches of SpaceX Falcon 9 and over 7 commercial launches of SpaceX Starship delivering V3 Starlink satellites will add over 800 TBps of capacity. This will triple the capacity in orbit by the end of 2025.
The number of high bandwidth Starlink subscribers will increase from 4.6 million today to over 12 million customers. There will be over 10 million people using and paying for the direct to cellphone texting and other services.
There will be more Starshield military satellites.
SpaceX revenue for 2025 will surpass $30 billion. This will be more than the budget of NASA.
Quantum Computers
There is robust error detection in quantum computers. Quantum Circuits has proven error detection.
The Equivalent of Built-in Checkbits
In conventional computer memory there are checkbits to detect errors. Checkbits in computer memory refer to additional bits used in error detection and correction schemes, particularly in the context of Error-Correcting Codes.
Their superconducting system uses cylindrical chambers with reflective mirrors inside to hold a photon. They use RF frequencies to control the movement and actions of the photon. The different RF frequencies determine if the photon can pass between the chambers. Currently, the chambers each have a diameter of 5 millimeters. They wlll shrink those down to microscopic size in later generations.
His work in this area includes: Developing quantum error correction conditions, also known as the Knill-Laflamme conditions, which provide criteria for quantum codes to correct errors
Advancing topological quantum error correction, which uses topological properties of quantum states to protect quantum information.
He describes the advancement beyond NISQ. It should be 4 years plus or minus 2 years to achieve quantum computers with very useful and valuable scale of millions of operations.
Quantum computers will be used for large scale and highly complex simulation, chemistry or optimization problems. They will not replace GPUs for AI.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Hey, Brian, this has been my go-to tech site for forever and I always appreciated you covering Lawrenceville Plasma Physics back when I was forwarding you their newsletters. I had a less than ideal departure from that fusion effort in 2014 after I raised serious corporate governance concerns with the investors, but it still deserved attention if only to point out the need for improved business and engineering strategies–More Elon-style, even if he drives me crazy most days! In 2012 E did tell me politely he “didn’t have the bandwidth for fusion” at the Mars Society banquet timed for the Curiosity rover landing, his date ended up wearing the copper gasket I had brough from our New Jersey dense plasma focus vacuum chamber as a necklace ;-D
SkyTran personal rapid transit was another inspiring idea I tried to nudge forward that shows how hard it is to truly innovate, unfortunately I have seen on Wikipedia that their latest backer has closed them down as of 2023, although all the references there seem to pre-date that. The fundamental Inductrack patents from Lawrence Livermore have all expired and so no longer need to be licensed, and similarly the focus fusion company’s core patent will expire in 2027, so perhaps open sourcing and new AI-powered innovation (and ever growing AI energy demands) will revive some of these in a way that is compatible with a Boring Tunnel/autonomous robotaxi/Starship point-to-point world?
& whatever happened to Broad Construction in China?
One last tip, Brian, when you screenshot a PowerPoint to make a graphic, go into presentation mode first to avoid the red spellcheck squiggle! ;-D
Brian, what share price for Tesla do you imagine now? Still 500 then 1000?
The miss on cars while disappointing is not a big deal. They have to transition to a new model Y in a high interest rate environment.
TSLA shares being down $100 from the new all time highs is a good buy the dip opportunity. Earnings on Jan 29 will be good. There was 11 GWh of energy sales at 30% margin. Each 5K of cars was $31M less net income. Tesla reported a record energy deployment of 11.4 GHw for Q4 — this was about 5.2 GHw higher than the average of the last four quarter deployments.
If the average gross profit per GWh is $105 million, then 5.2 GWh x $105 million = $546 million in incremental gross profit (versus the last four quarters). New Shanghai megapack factory started and it will be able to produce 20 GWh per quarter when fully ramped. $2B per quarter in incremental profits. New Model Y rolling out. The earnings call will have good car margins. Only 460k cars made vs 495K sold.
Tesla can still get back to $500 but it would be mostly in the last days of the month with the earnings call. Shares could easily hang in the $380-440 level
Need actually important events to happen, which we do not know exactly when they happen. FSD approval in China. FSD approval in Europe. FSD Licensing. The ramp of the new Model Y. Announcing the new model cars likely March to May. FSD improvement and getting robotaxi
If I may chime in again? What I used to do, with my first company (in the mid 1980s) was to mine the scientific journals. Before the internet, we had paper, books and magazines (scientific journals). No really, it was slow, but in your hand! Very hard for anyone to say, “they didn’t say something” when your holding a hard copy, in your hand. OK, these days, you can fake almost anything. How to verify? Get your information from multiple sources, ideally at different times. Oddly (or not) staggering the time you get info from different sources, lets you judge the “efficacy” of the information.
OK, lets go back a few decades to understand what I mean (sorry gang, I had a stroke 3years ago, and it’s VERY hard to type) But anyway… Historically, the first source to report “the news” was considered “the best” It might be inaccurate, incomplete, or just f****** wrong. But if it was first, that’s what you heard, and you remembered. The old news organizations KNEW this. The “power” of being first, had nothing to do with being “correct” but it’s what most people “remember” Honestly, I consider this an amazing 20th century sociological affact. I’m tired, more later.
I love this site. Why? It deals with two areas easy for me to deal with: Science and technology. But occasionally it deals with who and how the people “that” affects. So, I chime in on certain “social issues”. Science and technology are, IMO, the best way to change the world for the better. Oh, it just doesn’t seem to work out that way. So many have tried, over centuries. Do you think “better technologies, better products”, etc. will change the nature of our world? I used to think it might. Today? Nope.
Self-correcting machines, self cleaning surfaces, and other technologies are coming. that will blow your minds. How do I know? I’m one of many working on such stuff. IT WILL HAPPEN. Point? Our technology changes very quickly. Our nature? A lot more slowly.
Yes, like several US states banning lab grown meat before its even ready for public consumption. Sad, but true.
Brian, while i do appreciate your work, over the years, here on NBF, you promoted ECat, EMdrive, various fusion startups that should have already delivered GWH of electricity, fusion tech for space exploration, not to mention the posts that are paid advertising (there is nothing wrong with that) and the almost weekly references to how the next version of autopilot (now full self drive SUPERVISED) will be a dealbreaker. Either you have a very narrow definition of prediction, you constantly revise your predictions to match the data (like posting once that ECat is a fraud after posting dozens of posts considering it a valid scientific breakthrough) , or you are not close to 90% accuracy (which would be perfectly fine, but i understand that does not sound soo good)
Regards.
Could it be that E-Cat is not fraud and could change the world? https://e-catworld.com/2024/12/30/i-am-transmitting-to-the-new-partner-all-my-knowledge-andrea-rossi-interview/
No, there is no conceivable universe where E-Cat is not a fraud.
I know that the universe is under no obligation to be plausible, but even the universe shrinks from introducing world changing technological advances via con-men with a history of fraudulent world changing technological advances.
35,000+ articles. If you find over 3500 articles with issues around predictions, then let me know.
Here is a list of public predictions that resolved in 2024.
https://www.metaculus.com/contributions/baseline/110947/?year=2024&duration=1
Here is 5 years worth of prediction results
https://www.metaculus.com/contributions/baseline/110947/?year=2021&duration=5
My writing that emdrive has possibility of working and that nuclear fusion startups could succeed is not predictions with specific claims.
The nuclear fusion companies are still getting funded and they could succeed. They are generating nuclear reactions.
6 minutes into this video I say that none of the nuclear fusion startups or projects have lit up light bulbs like the 1951 nuclear fission reactor.
https://www.youtube.com/watch?v=c0FvFE778Kc
A prediction on the public platforms is a statement about a future event or outcome that specifies:
The Event – A clear, unambiguous description of what will happen, including all relevant parameters (e.g., “The temperature in Tokyo on July 1, 2025, will exceed 30°C”).
The Resolution Criteria – Explicit, objective criteria by which the prediction will be judged as true or false. This includes:
Source of Truth – A reliable, publicly accessible source or method for verifying the outcome after the event (e.g., “Data from the Japan Meteorological Agency’s official records”).
Time Frame – A specific date or time period by which the outcome must be resolved (e.g., “The measurement must be taken at exactly 12:00 PM JST on July 1, 2025”).
Falsifiability – The prediction must be structured such that it can be definitively proven true or false based on the resolution criteria.
This means:
The criteria must not be open to interpretation.
There should be no ambiguity in how the outcome is measured or reported.