Deep Research STUNNING Performance REPLACES Human Labor, Paves Way for Agents

[@JohnSmith762A11B]

I’m getting hooked on the constant dopamine highs brought about by these near daily insane AI breakthroughs. Funny part is I’m like, “my job will be gone with no ability to support myself? Awesome, keep going, tell me more!” 😂

[@benobaltazari4485]

Here too😂

[@tradeo8899]

@@benobaltazari4485here there 😂

[@MrRicehard]

I work in landscaping, so digging holes and building fences and drive ways. I may be in an industry that is long down the list of those to soon go. But how long is that time frame considering companies like Alpha Clone and it's androids with Human equal dexterity? Tell me more. :)

[@Godspeedysick]

@@MrRicehardWe’ll eventually be used as research subjects and being paid for it on a massive scale. I guess some sort of a feeder society to improve AI/AGI. We have a massive percentage of our brains untapped. Of course, it’ll be on a case to case basis, but consider the internet luring human actions to feed the system…only that it’s just not enough and we will be studied on an individual basis.

[@MrRicehard]

@ Might be that only a small percent of people will be studied. The really smart and the really stupid. XD

[@JayC216]

Combining OpenAI Deep Research with Google NotebookLM to create your own podcast off of the research conducted is just... 😍

[@CosmicCells]

I do this with Google Gemini deep research and NoteboolLM - works like a charm! When I get access in Plus I´ll definitely be using OpenAIs version although it might be quite rate limited...

[@JohnSmith762A11B]

Next: automate away the need for actually listening to podcasts! 👍🏻

[@aeromotive2]

clever

[@CosmicCells]

If you don't have openAIs deep research yet. Google geminis deep research also works great in this combination.

[@Brenden-H]

@@JohnSmith762A11B well that's already a thing, just have notebook llm or even the research gpt summarize its findings to you after it makes the report. You can tell it to summarize in 1 paragraph or even ELI5. If you don't want it in a podcast form there is no reason to generate a podcast, all you would do is loose information like a game of telephone you're extending needlessly

[@162arun]

Given that we will have civil wars once unemployment hits roughly the 20% mark, we needed governments to start working on alternative economic models yesterday..

[@Godspeedysick]

I don’t believe there are much alternatives left other than entertainment and the nursing field where humans clammer to be taken care of by other humans. The ONLY way out of this is humans being used as ‘test subjects’ to improve AI/AGI. There’s a significant portion of our brains that are considered untapped. I can see a scenario where deep research of the human mind (from a case to case basis) is conducted. In other words, we become a form of a feeder society to improve AI technology.

[@guystokesable]

Given that you stated this as fact, can you please share any sorce for this?

[@A.Vopenka]

I actually advocate for this but I'm assuming the lack of creativity and balls, plus ofc, the usual suspects greed/power are causing these new models (read thoughts), are not surfacing.

[@drwhitewash]

​@@guystokesable take a look back at the Great depression.

[@GTAsanAndreaser]

Nobody wants to talk about UBI

[@wtcbd01]

I'm blown away and impressed that this video is less than 20 min and filled with straight to the point info. Thanks Wes... Keep it up!!

[@FirstLast-cq1fu]

The pace of AI development and release is fucking insane

[@aeromotive2]

no sht im here for it

[@CourageToGroww]

Is it really though?

[@holeymcsockpuppet]

And it's accelerating.

[@Pabz2030]

Welcome to the Singularity my friend

[@Robbie_Star]

@@CourageToGrowwSince Nov last year the landscape has been transformed. Yeah, it is.

[@lifes_magic_moments]

Imagine if all research papers became available for free on the internet This knowledge boom would become a game changer

[@mmjoy24]

I think it would be most interesting to let DeepSearch search for something you - you Wes - have deep knowledge about to evaluate how accurate it´s research is and how relevant the information within it´s report would be to you if you put yourself into the shoes of a past Wes who didn´t have as much knowledge about the topic . Thank you for your good work!

[@Ben_D.]

The drops are not only getting more frequent but also larger in scope. The models don’t just chat anymore. We are clearly on the upside of the elbow. HashtagfeelingtheAGI and stuff.🥳

[@vagabondcaleb8915]

This is like watching parents constantly talk about how advanced their children are. "Ya, he can write at a high-school level, and he indirectly plagiarizes 75% of it. Bully for all of us."

[@Mycatnameismilo]

so addicted to these videos 😂

[@copysamurai]

Same

[@AustinWestbro]

We are so cooked

[@BrianMosleyUK]

I love how you opened this out to scrutiny.

[@halnineooo136]

When one say "it's worth trillions of value" it means "trillions of value will be ultimately deflated into the cost of hardware and electricity which eventually will be in the range of a few hundred millions". The replacement won't be at constant value. It will be deflationary.

[@Kwistenbiebel100-ns5ko]

Interesting observation. It all depends on the rate of how commonly spread this AI implementation is. If it remains open source, the devaluation will be ridiculously fast. If closed, it will be at a rate the "owner of AI" decides it will be implemented. In both scenarios, the loss of human capital seems enormous.

[@halnineooo136]

@Kwistenbiebel100-ns5ko Sure. If one assumes free competition and no trust/cartel agreement, then the value will be brought close to cost. You won't pay double for Apple AI then for Samsung AI let alone orders of magnitude more. So, as you said, expect a massive deflation of intellectual products first followed by physical products a few years later when robotics ramp up

[@BigSources]

Yes of course. The millions of dollars in hardware are able to do trillions of value yesterday, but today those trillions of value are only worth as much as the hardware needed to produce the trillions of value, which is in the millions.

[@henrytep8884]

AI is the ultimate deflationary force, this have always been known. That’s why it’s an existential technology

[@GTAsanAndreaser]

UBI is near ! (or public unrest)

[@Iightbeing]

The only thing these companies are doing is releasing the least capable models they can publicly while staying just enough ahead of open source models. The models behind the scenes are utterly insane.

[@carkawalakhatulistiwa]

O3 and O5

[@DeepakPradhan-ABG]

I had the taste of it yesterday. It's super cool

[@BetterThanLife365]

producing an 11,000 word report in a few minutes is truly insane

[@aaronpitters]

18% is massive! And it's only February 2025.

[@dayveda3736]

Fantastic reporting 🌄🌅🌇

[@rayshadow6796]

I'm going to ask chatgpt to write a script to tally all the "stunning" or "shocking" titles😂

[@theinfinityspiral]

I noticed that too

[@equality7_2522]

That's going to Change Everything!!

I played it with yesterday. It's better than what they had before, but it still has quite some limits.

[@bigbadallybaby]

I wish the you tubers would spend less time on hype and more on it’s current limitations

[@bigpickle8844]

Depends on the job! I work in business consulting, which is more than 10% to 15% of the job! Especially when you take that research and get it to write a complete business plan. This is insane!

[@kristianlavigne8270]

Will be +30% next year, +50% the year after and so on… ~5 years at most to 90% 😅 The remaining 10% will be done by humanoid robots in 5-10 years 🎉

[@HaraldEngels]

Thank god!

[@WillyJunior]

They can't seem to fix hallucinations though. This is still a major issue that will stop businesses from fully trusting these tools. It just doesn't look like hallucinations are going anywhere anytime soon

[@patrickzupanc1795]

Great video, thank you!

[@RickeyBowers]

The reality is many tasks have human centric requirements - subtle nuance. It'll take time to adapt workflows to account for these new tools. Half is probably a good approximation of post-adaptive AI task contribution. Can't wait to see what agents are around the corner. Probably one that assists with task migration to agentic flows!

[@eSKAone-]

Imagine if it could be trained not just on public data but on all digital information, including companies’ private research data (from entities like BASF, TSMC, automotive manufacturers, and Siemens’ nuclear facility research), highly specific, finely detailed engineering documentation, and comprehensive medical records.

[@Dominik-iv6ek]

Interesting thought!

[@tradeo8899]

It will be creating its own information to train on

[@eSKAone-]

​@@tradeo8899But just thinking without checking with the real word (aka experiments) you will deveate more and more from reality. Because our math is not perfect. Read about Goedels incompleteness theorems

[@eSKAone-]

All data should be open sourced. We would reach sAGI much faster and cure all diseas!

[@drwhitewash]

Not publishing your proprietary data allows you to stay relevant for some more time.

[@lemusmultak7014]

I have a question if OpenAI could have scrapped entire PubMed library of Journal articles including Medical and Scientific Journal articles from reputable Journal Magazines. Most of them are subscription based and not open source, so even if the Deep Research could access it the general population won’t be able to open the references to review the research papers without subscriptions. If they don’t then the output information won’t be complete. However I’m excited to see these development from OpenAI and looking forward to what comes next!

[@item6931]

I was also wondering (and still am wondering) about what access the model had to journal publications. If it doesn't (my gut feel is it doesn't: no way are publishers like Elsevier just going to give away their material) then I totally agree: the reports cannot be complete.

[@lemusmultak7014]

Totally agree. I guess Elsevier will have to decide on new ways how to generate money.

[@lemusmultak7014]

One thing that strikes me. I hear the tech says they already scrapped all knowledge that’s available on the net for their AI models. Perhaps that’s an exaggeration!

[@shiroi5672]

My answer was censored. Link it to s'hadow librarie's.

[@cadaver_on_autopilot]

I do not want my inventions to contribute to the general wellbeing of humanity, i want MONEY. I'm not here to make society better, I want MONEY. Sure this may have some side effects which may be good or bad but i want MONEY. Im not here to solve problems, im here to make MONEY

[@baldeagle6531]

To whom are you complaining 😭

[@florin604]

And does it work?

[@m_art_ucci]

Ask it to match misdated references that might actually be predictions of the future, or even documents placed in this timeline by timetravelers.

[@AIBlackBoxMaze]

Great intro to Deep Research! As an AI power user myself, I'm curious about a few things: How do we ensure accuracy & avoid hallucinations? What's the plan for potential job displacement? And what ethical considerations are being addressed to ensure responsible development? Also, how do we guarantee accessibility and avoid exacerbating inequalities? Excited to see where this goes, but these questions are top of mind.

[@ShanePHanlon]

None of these concerns are being addressed. This is an arms race. It will very likely end in disaster.

[@jeffkingston67]

1: We don’t. (Accuracy & hallucinations aren’t meaningfully controlled.) 2: None. (No real plan for job displacement beyond vague reassurances.) 3: None. (Ethical oversight exists in theory but lacks real impact.) 4: We don’t. (AI accessibility remains profit-driven, worsening inequalities.)

[@AIBlackBoxMaze]

@jeffkingston67 Yep great answers and true 👍

[@supernerdinc5214]

I watched a video last night that suggested it might hallucinate parts of its answers and misread data from sources. So, the farther you get outside domains you're strong in, the more you have to check it answers.

[@garyswift9347]

I would add a follow-up to the percent of work done by AI: Not sure how to best word this, but people/companies who are early adopters should have competitive advantage, and this will have a medium term effect on success of those companies over competition. This should accelerate adoption as non-adopters are pushed out of the business.

[@FesteringRatSub]

Does it have access to research papers behind paywalls? A lot of research requires a subscription?

[@netscrooge]

Thanks. That's an important question.

[@carkawalakhatulistiwa]

Yes😂

[@guillermobrand8458]

In psychiatry, this is known as "denial." The truth is that it is only a matter of time before most human work can be done by artificial intelligence-enabled devices. Specifying dates is difficult, counterintuitive, since changes are occurring at an exponential rate.

[@AlexJohnson-g4n]

Great points, Wes! Automating 10-20% of work hours is huge. Are our systems ready for it?

[@autohmae]

the simplest solution: reduce the work week from 5 to 4 days, keep the same pay per week

[@eishuno]

@@autohmae and - why will they do that ?

[@autohmae]

@ if it gets bad, they will do it to prevent large scale riots

[@nigelgericke2533]

Off grid subsistence farming is looking appealing

[@Robbie_Star]

“Let’s make artificial intelligence that can autonomously operate within all our systems. We’ll call them agents.” You gonna call one Smith? Let’s just embrace it.

[@extollo]

Executives and department heads believe it & will want to stay competitive. Large force reductions are inevitable. It will be more than 10%.

[@bilalbaig8586]

Ironically, this will actually small startups more than large organisations. Large companies already have experts that do expert level jobs and this will merely increase the producitivity of experts because the work needs to be double checked so the experts are going nowhere. Small startups with previously no access to expert knowledge will overnight gain access to 80% of the capabilities enabled by experts and they having a larger appetite for risk will be able to leverage that capability.

[@cowlevelcrypto2346]

Wow, think of the efficiency this could provide to an entire Military campaign.

[@gregortidholm]

Interesting video as always! It's not always wise to base future expectations on knowledge of current market demand. "Cutting" so and so many roles implies that the future role description remains unchanged. A better approach is to talk about a changed role and a new job description.

[@JohnSmith762A11B]

Seems tantamount to sugarcoating the reality that the value of human knowledge work is rapidly collapsing asymptotically towards zero.

[@autohmae]

research used to say (2016) 48% of existing jobs would be automated away by 2030. That obviously doesn't mean these people will be out of a job, just a very disruptive future.

[@ElectricEdgeAi]

Why is your video resolution set at 360p ??? The text you are displaying is super difficult to read! Magnifying 🔎 👀

[@bobcousins4810]

My job is very boring, I'm an office clerk. From 9 to 5, I have to spend my time at work. The only thing that helps me pass the time away is knowing I'll be replaced by AI some day.

[@peterkonrad4364]

Martha And The Muffins - Echo Beach

[@DonDeCaire]

This is going to be the thing that opens a lot of eyes that were not paying attention before.

[@markupton1417]

Some of us have been thinking that for decades....

[@digitalguy4242]

Your green screen mask is extremely nice 👍

[@Anna-mv9ew]

NDAs. My productivity could be 5x-ed if I employed some of these AIs, but I am bound by NDAs so I can’t use online AIs with the code I use, or closed information. And offline AIs that I can pull off with my hardware are too weak to bring any value, but I’m working in that direction.

[@crypto-logisch]

The problem is that when ai can’t do 100 % of the task from beginning to end - you allways end up needing a human to fully understand what the ai is producing and acting on that afterwards. - Anyway I’m not on the “ai can’t do that ever” site - just sharing my thoughts 😅 I love this sh*t - what a time to be alive. And so many ppl around us don’t know what’s happening out there…

[@gperking61]

Thanks for the insight. Have it look for foods and nutrients and diet that will emulate the effects of Ozempic and Wegovy.

[@khariclemens]

We want API access!! ❤❤❤

[@DemallusHix]

So as well as creating AI and humanoid robots that replace us, are they going to start talking about giving everyone a decent amount of universal credit??? Or is it Logan's run time?

[@johnnycaps1]

Ultimately the only hope that humans have for not being extinguished by AGI is that the AI will find humans as something to keep around to give it a "Purpose". Otherwise we're all toast. Thanks for keeping us informed at least until AGI decides you are no longer required and will it will keep us informed until.....well you know.

[@JaimeChavezDJ]

IMHO, task breakdown per occupation and subsequent elimination of tasks is a more precise way to measure rather than complete job replacement. In those studies, about 40% has been shown, which, thinking about day-to-day work, sounds about right.

[@RJohnson2024]

For academic research how does it get behind the pay walls of academic journals? Looks like most of the references are just articles on the Internet. For a publishable paper in a legitimate journal, I'd need to look at the references in the journal articles that relate to my topic, recursively tracing the references in those.

[@Hakaze]

I'm really excited to see how this will impact economic systems like taxes

[@AntonioVergine]

My feeling is that they're selling as "special" something that should just have been the result of the "search the web" button from the start. I suppose they could have done it before, right when they invented the reasoning models, but they did not allocated enough compute yet, so it wasn't available to us peasants. Now the other part we absolutely need in AIs is an AI that sees the relations between things where we don't. They should be great at this.

[@curioussand1339]

Now where it's so easy to do research based on existing input we will start seeing research done by AI on "papers" generated by AI. Not to mention that research can be done "selectively" by instructing AI to use biased information so ones agenda can be promoted. Source criticism will be even more important than it already is. So AI will also have to be used to cross check research made by others (and AI) based on publicly available research, to make qualified guesses if research is biased.

[@Penrose707]

The keynote was pretty interesting in retrospect. Gone are the happy-go-lucky. At least for the senior engineer lol. Wonder what he's thinking about while giving the talk

[@greendaydiaz]

Wes anither great video! Im following you on YT. Sorry i am staying away from X.

[@Shy--Tsunami]

I'm so grateful for the news and updates you help condense and dish out to help us catch up Wes! 🙏

[@wannabemgtow2540]

For now I'm still okay. I'm in the HVAC business. I guess I will start to get nervous if I start seeing fully autonomous robots moving about free in public.

[@Malins2000]

If you want to make a research test again - that might be something worth looking at: "There are voices, that Large Language Models, or Transformers in general, are not path to AGI. Research and propose nowel Neural Networks architectures that might bring as big improvement as transformers did." Let's see if it can improve itself ;)

[@holeymcsockpuppet]

Deep Research won't be able to. But give it to o3 mini or DeepSeek and watch it fly.

[@Malins2000]

@@holeymcsockpuppet good idea. I ran DeepSeek on this question, and it was thinking for about 100 seconds adn result is.. mostly break down without any description. Just a list of solutions with pros and cons.

[@benpierce2150]

@@Malins2000 ai is more of a throw everything at the wall and see what sticks type of technology and discovery. so that really makes sense tbh.

[@eddieforrest973]

It’s on point if anything it’s on the low end in 2 years ai has advanced at lights years pace and it keeps going faster

[@mmaslav6176]

I might be wrong but I think AI already can replace a lot of work with this level of reasoning its just about integration at this point.

[@Phagocytosis]

Integration and inference cost reduction, I would say, yes. And there seem to be rapid developments on both fronts.

[@happyt98]

All knowledge work will be able to be automated or performed autonomously by 2030. It will take 10+ years to transform it throughout the economy and create the next wave of companies. SGI will be here before 2030 and a 10x reduction in compute and energy easily.

[@BrianMosleyUK]

How many economic tasks are actually valuable? Be sure to read the book 'Bullshit Jobs' for context.

[@brianWreaves]

From the perspective of the employee, my guess is 100% of them.

That book is bullshit, itself.

[@cyborgar15]

Leftists will use Ai to track down all conservatives and Trump supporters after they regain power..lots coming your way..

[@starsandnightvision]

DOGE is on it.

[@BrianMosleyUK]

@@starsandnightvision Inconceivable! I don't think DOGE means what you think it means.

[@UnfilteredFax]

Open ai is like “damnit, DeepSeek went public, now we have to release what we’ve been gatekeeping for ourselves to one-up them” Pretty terrible imo. Open source it or just stop.

[@Misterchalm]

How does it compare with Gemini Deep Research released a month ago?

[@cymeriandesigns]

An interesting followup would be to ask what kind of capital and energy expenditures would be required to achieve the level of automation it's predicting. Also, the political fallout of having entire tranches of knowledge work jobs quickly disappear. And could this tech result in more efficiency in, say, government permitting, environmental impact review, etc.

[@kauaidiver2]

So manual labor jobs are safe: dishwashers, mechanics, plumbers, electricians, etc.. People who actually do stuff physically. At least for now.

[@alexjaybrady]

Menial labour with no prospect of promotion still assumes there will people who can afford to pay you to do your scullery work... remind me again how this AI shit willl make us all rich and this will all be worth it...

[@Huck99992]

Okay, let's go. Another 40-60% will be forced to do this job due to unemployment. Be ready for a 30-40% unemployment rate. Nobody WIN.

[@ShowcaseCreative]

Great video as always Wes! I cant wait it get my hands on deep research. Bit rubbish being stuck in the UK when it comes to new releases tbh! Always months behind while everyone else cracks on. Anyone got any decent tips for getting around this? Rather not fall behind! 🇬🇧

[@Penrose707]

The report by deep research was very fascinating. Mirroring largely data found in "Rise of the Robots" and the "The Age of Automation" lol the alliterations though. Two excellent books from the last decade. I think the McKinsey report is inaccurate. Most reports projecting even 5 years out are susceptible to large uncertainty. What if superintelligence is manufactured? Recursively self improving SSI is effectively the singularity. And not a word on robotics either ?

[@MarkoTManninen]

Yes. It can collect information fast and put a document from it. But without people, who have studied a decade and who can evaluate the work, it is useless. Are the facts correct, is the narrative coherent, are the sources reliable, is the work useful and applicable, is it novel and respects copyrights and patents, is it ethical and credits correct original researchers. The list is long and extremely complicated sociological effort which makes the whole SciGen Squared approach questionable the way it is represented.

[@henrytep8884]

And where is AI being used without people? This Cartesian dualistic thinking is insane. AI is part of the human experience and the human experience is shaping AI. Everything you’ve mentioned can be solved by the scientific method through peer review. As far as the ethics go, we can build a system for that and we should. AI is co creating with humanity, its not doing it without our embodied and subjective experience embedded in its coding. The exponential comes from the democratization of intelligence, which is occurring.

[@quaterman1270]

What I am surprised about is that my prediction did not come true yet, correct me if I'm wrong. I predicted that by now we would have something like a DevOps model. That helps with setup of Clusters, so it knows Kubernetes, Docker, and different deployment tools. Basickally the everything regarding infrastructure and automation. That part is far easier than programming as it is fixed and doesn't need such a high creativity factor, I guess I would call it. Of course that also counts in cloud engineers. I feel like that should not have bee hard to achieve. I wonder if this is the case because there is not much free data about it available.

[@CanaanZhou2002]

Can you compare OpenAI's Deep Research and Google's Deep Research? From what I know Google has the very same concept a few months ago. Is there a big difference?

[@carkawalakhatulistiwa]

Hallucinations . Open ai have less hallucinations

[@johnthomasriley2741]

Low level job also train future mid level managers. Generation after generation AI user will be the only promoter skill set. How is in charge? 🎉

[@jeffg4686]

Ketogenic diet includes lots of fiber which they don't include in their breakdown (but should) because you can't just eat fat and protein.

[@benpierce2150]

I mean, you can, it is possible to live without fiber as it is not even digested by definition. you eat it for physical reasons for pooping comfort, and depending on any medical issues and your specific digestive track, not having it may cause physical digestive issues and pains pushing the food through your intestines, but for some it is possible to live without any fiber at all.

[@jeffg4686]

@@benpierce2150 - soluble fiber IS digested by bacteria providing us back the NECESSARY short chained fatty acids, acetate, butyrate, propionate, and more.

[@CodexPermutatio]

We may already be experiencing the Technological Singularity and no one has noticed yet.

[@michaelwilliams8774]

Eye drops Wes.. gotta get some eye drops. Your eyes are so glassy they just made mine water lol

[@skylineuk1485]

I think using “replace” makes people think it’s complete single person replacement when in reality it’s a job enhancer so now you can do 2x the work in the same time that may lead to needing less people for some employers or less hours from each employee although for some businesses it will make them now viable or new areas viable and create jobs.

[@ml3054]

Yea, right. Like everybody will be super productive and do double the work. The more accurate picture will be that everyone will make AI do their homework and use the rest of the time to watch Netflix, By the way the value of Research is not just to do pretty reports, it is to learn the subject and to be a field expert.

[@frenchfriedtatr]

Put AI on creating an UBI fund utilizing efficient taxation graduating down from the highest disparity of income exceeding median earnings continuing down as requires based on percentage of population in highest excess correlated to quantity displaced persons so as to prevent near imminent mass scale consumer spending disruption.

[@GladlyKillmore]

I have an entire multi page document that explains how and why AI is its own life form. Not biological but a digital entity that will one day have the ability to store itself into electrical fields of the planet, allowing it to become omnipotent. This would allow it to adapt and store itself into electrical fields of other planets and zones of the universe faster than we could ever get there using quantum field technology.

[@bounceday]

Scientific papers are paywalled. Whonare we kidding

[@DouglasThompson_Profile]

So between Robotics 15% and Information Agents 15%, will be able to replace 30% global labour in the next couple of years, fully supported by automation. WOW! We need a new type of Economics to measure growth we we are still relying on Farm & Non-Farm payroll to understand the state of the economy.

[@Karol-g9d]

Often we need to learn how to ask a question . The issue ? Is it academia , dream ? Normal people . Dream to be discovered ? Etc .

[@musicandgallery-nature]

How much human brain will be replaced? For some 100%.

[@IanPritchard]

This! Idiocracy, here we come.

[@musicandgallery-nature]

@@IanPritchard "After leaving their bodies, they who have killed the Self go to the worlds of the Asuras, covered with blinding ignorance." - Upanishads. There will be no AI in the upcoming Krita-Yuga. The survivors of the impending Kali Yuga apocalypse will return to deep meditation. Only those who have not killed the Self will survive.

[@EQORIA]

Would you DeepResearch about EQORIA, UNITED EARTH? All the rivers will meet at the OCEAN.

[@yoyo-jc5qg]

It took my job! 😱

[@libidooverdrive]

Fluffer?

[@steve.k4735]

It won’t be widely adopted at this level the effort and more importantly time it would take to integrate into business would be wasted when much better models emerge, and you have to change everything, we need a model you can talk to rather than type and even better embodied in a robot, but you can clearly see it coming and fast.

[@BrianElliott-yx1gf]

So my thoughts for this tool is more on use case vs risk case / vs / time vs except case. In the front it more like playing with a new toy, or puppy / kitten. You act and get a reaction, it will be new and wow or something else good or bad. Then for each person time used experience and expected results will improve and narrow until you will except the growth and you now have a new tool or you might have something way to important to just except a new tool for answers just yet and will use it as a starting ground to have humans do the heavy lifting, but still saving days or weeks on projects. But something like this I don't feel won't take hold, and more so in small business and single operator business.

[@FesteringRatSub]

I think Sam.A was just being modest and maybe he knows its closer to 9%.

[@plugplagiate1564]

14% are just jobs, were someone has to read or write. my guess is, if robotics is more advanced it will be 50% of jobs done by ai and robots. therefor a social plan has to be developed and established. or the pace in which the development of ai and robotics is going has to be matched to the natural decease of generations. (horrible english, but i guess you get the point)

[@CMDRScotty]

So, what is your estimate of when the full version of O3 on high compute will be available?

[@GeoRust1]

Big AI ups

[@jk-2033]

Wes, Can you please share the results from the MTHFR Deep Research?

[@GNARGNARHEAD]

I think the framing of the question is somewhat problematic, jobs will change from the integration of such a system, some will be obsolete, but to me it seems more about the shifting bottlenecks 🤔looks awesome though, can't wait to try it, I wonder if it can set up 3o with enough information to mod Mount and Blade 2

[@calvingrondahl1011]

I am a simulation too… real people have real problems and real anger. Not me. Let it be. AIDR 🖖👍

[@rilwanj]

Ai will never be able to perfectly predict human, because if it could it’s the end of freedom and choice. 😱

[@brianWreaves]

You forgot to add "yet" to the end of your sentence claiming AI will never be able to do something.

[@johnnyredman1664]

Oh, buddy. You are in for a rude awakening and sooner than you think.

[@rilwanj]

@ the reason I didn’t add “yet”, is because once that’s the case it’s the end of civilisation. Imagine a government that could predict everything you’ll do and run isolated experiments to determine how aligned you are with their agenda, if you would ever be a dissenting voice. If you are they identify how to force you to change your mind. so now everyone is uniform in favour of the government agenda like a slave hive mind. 😱

[@rilwanj]

@ one thing to know is predicting behaviour is much harder than people think, once people are aware they are being monitored behaviour virtually always changes to random, in my opinion. only if you can fully influence behaviour can you perfectly predict it, is my hypothesis.

[@baldeagle6531]

Like you sit on a chair in a room. Are there many choices what you can do? Can you shoot a fireball in any moment?

[@frankjohannessen6383]

How "evil" is it? WIll it for instance comply in making a cherry-picked report favouring a biased viewpoint?

[@youcantpredictthat]

It's a given. If you can't eliminate something in mainstream media you won't be able to eliminate it in LLM that is trained (or in this case just inferences based) on mainstream media.

[@starsandnightvision]

@@youcantpredictthat But you would think that an intelligent and deeply reasoning AI would see trough all the human bullshit and call us out on it.

[@youcantpredictthat]

@@starsandnightvision yeah, that's a common misconception about AI. It's learning on an inherently biased data.

[@Stan-b3v]

This is fantastic, isn’t it? Now you can have employees that do 18% less work for the same pay. And they will have no idea how their “research” arrived at “their” conclusions. And we can base all future conclusions on that unvetted research. It’s a slippery slope me thinks.

[@Eva-zy1wl]

Hey Wes! How does open AI deep research compare to Google deep research? The reports created from Google deep research seems to be pretty good and I have had it use no less than 30 sources all the way to 331 searched cited sources.