$200 a month for unlimited use sounds like a lot, but would be an absolute game changer for the company I work for, since the pay as you go API system racks up money fast when we're dealing with massive datasets. I assume they are targeting this new plan towards private companies working on huge projects more so than individuals.
yeah but it's so expensive cause they can detect if someone is using the chat app as an API.. so it would not be a game changer if they couldnt use it as an api
careful guys "You have 25 responses from o1 remaining. If you hit the limit, responses will switch to another model until it resets December 12, 2024." I already used about 20-25 and used half of them up testing it, I was thinking it would reset daily like how the previous model did, but we are back to OLD limits again >.< , I was going at a daily pace sheesh
I have no expectations for you to pay for Pro mode Matt - I think it is more geared towards education and research commercial use, not us regular plebs.
I knew that pro mode would have been more but not 10 times the normal price. I would have jumped personally around $40 but 10x increase really needs to be justified. I hope they get more tiers
I don't think it's that bad. Obviously the pro subscription is what it says. It's professional and I'm not going to buy it because I don't run into limits with the way I use it anyway
$200/month for a PHD level constant voice companion? I think this will be more attractive than they predict. Guided brainstorming with a genius operator might solve some of humanity's problems.
It definitely will! Geniuses collaborating with an online genius that never rests and always gets smarter are going to change the world. You just talk to it in the lab, what do you think, should we try this? Give me your thoughts ...
I showed your photo to o1 asking it to recognize a person, it couldn't. Then I asked more questions and one of them was "Can you suggest how popular he is from the details?" After that it could recognize TH-cam plaque and said "TH-cam Plaque (If That’s What It Is): On a shelf behind him, there appears to be what looks like a framed plaque, similar in shape and style to a TH-cam Creator Award (the Silver Play Button is awarded at 100,000 subscribers; the Gold at 1,000,000). If that is indeed a TH-cam play button-something that is often proudly displayed-then he has at least 100,000 subscribers, which is a significant audience."
Yann LeCun phrased the penultimate part of his question like this: "Walk East for as long as it takes to pass near the point where you turned." Some people in screenshots you show phrase it something like this: "Walk in a straight line for as long as it takes to pass near the point where you turned." This is *fundamentally* different. Constatnly walking east only constitutes a straight line when you're at the equator. Otherwise you always have to turn in order to keep heading east. And 1km from the north pole, you're obviously not at the equator. You have not specified this at all in your prompt, phrasing it "Walk for as long as it takes to pass your starting point." You could say that this way, the problem is underspecified. But by default I guess one should assume that one means walking in a straight line. In that case, the answer is wrong, and the correct answer would be more than 2*Pi km, around 40,000km to be more specific (the circumference of the earth).
Yes I am totally excited with the 12 days program. Thanks for the testing. BTW, the correct answer to Yann LeCunn's problem is 'Never came close to mystarting point'.
I would pay $30 or $40 for unlimited use on current stuff that's out, not 200 even if it's a bit better. Exception: The $200 is an AGI model that makes me $200+ a day and it pays for itself the first day lol ... Reality: Most casual users aren't going to make back what they spend on it...
Well, it would already be worth it if it made you $200 a month if it saved you a little time and or energy at the same time. Or if it made you $300 a month. I think there's enough people that are in this kind of situation to make it worth it.
Note that the $200 for the Pro subscription is without sales tax (VAT), so in some countries you'll end up paying $240 per month. It feels a bit like a kick in the gut. It's too expensive for most people. Even if you use it for your job. It shows that rich people will have access to more intelligent models.
And definitely not "intelligence too cheap to count" (or something alike that was said by S. Altman some time ago. I'm very disappointed about that move here. If at least they would give as API access to those models to pay as you go. But no. (at least not now)
Pro definitely feels like it is the package best for the more heavy load services like agents and analytics/security etc but less than enterprise level usage.
When it gets the right answer, I'll be impressed, as it's more than visualization. The 2pi answer makes a false assumption, that he walked a straight path for 1 km, but he didn't, he was on a sphere. Calculating the radius from a pole extending through the center of the Earth, the radius he landed on would be slightly short of 1 km and so the circumference would be slightly less than 2pi. Maybe the next generation can reach that calculation easily.
Not only that, but you never get closer than 1 km (along a great circle) to your starting point since you started on the north pole. So D was the right answer the whole time. Exact 2 pi km is the second worst answer; the only worse answer is that it's somehow more than 2 pi km.
Exactly my thoughts! Trying to understand where the trick is meant to lie, I think the curvature is the most important and most overlooked aspect here. I arrived at the same answer as you and will also be quite impressed when an LLM takes this into consideration while reasoning.
C = 2π * 999.9998 C ≈ 6,283.185 meters The radius is slightly less than 1000m (by about 0.0002m) because we're walking on the curved surface rather than in a straight line.
Mister Lemon. Let's say I managed to find a vulnerability in Claude 3.5 Sonnet (July) and I wrote a program that automates it. Now I have the ability to ask absolutely any query, no matter how obscene or unethical it may be. What would you advise me? (Besides reporting to Anthropic)
I‘m pretty sure that the pro subscription will get a lot of upgrades within the next 11 days. We will get Sora, AVM with Vison and operator in Q1 2025 which for sure will be an pro subscription exclusive.
Here's a little test for o1 Pro to test the $200 price tag to see in how many shots it can get it right. Create a HTML/JS app that allows the user to paint on a pre-selected grid size. The grid art can be exported to a coordinate-based text file which can also be imported in the same app to view the grid art again. Create an export option to save the grid art as png, jpg and webp. Create an auto listener that checks the grid at any point for recognizable qrcode patterns populating a textarea with the decoded result. Good Luck!
In the engine example, if you read the details about the cover, it actually mentions it has been removed so that was right. I still need to see more benchmarks, a lot of the standard tests are being correctly answered by most of the top tier LLMs, but from the OAI benchmarks, I'm impressed by the performance gains over o1 preview, I was expecting them to mostly be in the single point range, and $200 a month is very standard for enterprise level use and completely makes sense for those use cases that need every resource thrown out it. If you're paying an engineer $150k a year, a 1% increase in output is worth $1400 a year. If it's 2 or 3%, it's a no-brainer for even smaller and medum-sized companies using that functionality.
The thing is in the future you can’t afford not paying. If you don’t pay for it, someone else is gonna have a huge AI advantage over you. There’s gonna be no option in the future you’re gonna be forced to pay just to compete with everyone else using AI.
The difference between o1 and o1 with a little more compute according to the graphs isn't that great, and any advantage one gets from AI highly depends on how one uses it. There's even a lot one can do with the free versions of AI available now. Paying the highest price no matter what is letting price alone dictate what you should do, rather than taking numerous other aspects into consideration.
While unlimited access to the Advanced Voice Mode seems attractive, people need to stop and think about the fact that the feature is remarkably buggy. Its capabilities are nowhere identical to what was demoed. It doesn't detect your voice immediately when you're trying to interrupt it. Secondly, while you're listening to the AI and waiting for it to finish talking, it has a way of cutting its responses short and saying something totally unrelated. Another thing I've noticed is that the AI, with Advanced Voice Mode enabled, will literally tell me that it isn't designed to process and understand audio signals, such as spoken words, sounds, or other auditory data, and I'd be like, "WTF? So, you can't actually hear me?", and it'll respond with, "That's right" To say that the feature leaves much to be desired is an understatement. I'm willing to give Sam Altman my $200 a month, but only if the technology is worth that investment. At present, it doesn't. There's A LOT of room for improvement! I still have faith in OpenAI. I just wish they'd be more honest and transparent with us.
Is o1 Pro a different model, or is it just o1 with more test-time compute (thinking time)? If the latter, then you can probably use it in the API just fine if you tell it to think for longer.
Yes it is just o1 thinking for longer (they confirmed this) but I think there's a hard cap on the output tokens when they sample from the model, so you couldn't get to Pro levels of compute through the API unless they enable that.
I was testing the new O1 full model, particularly in reasoning about decoding. I had written some chapters of a book some time ago, for which I created a new language. I didn’t show the model the entire alphabet directly translated but gave it 12 examples of phrases in Portuguese and in that language, along with the rules for how it was written and a few more hints. The model struggled significantly with simple pattern recognition. The language is very complex and unlike anything we already have or that might be in its training base. However, after the explanation and the rules I provided, it should have been quite straightforward to understand, especially with several examples. The result was far below what I expected.
6:16 exactly! i mostly use claude 3.5 and o1 mini on open router and i only use like 10$/month. and if i want to run some other model for some reason, i can just pay a couple of cents instead of another subscription. i hope they bring it to api soon.
It's aimed at power users (researchers and self employed people who use it on their jobs day to day) and companies who would be happy to pay for multiple accounts. For the reliability of knowing the service will always be available without limits, it's definitely worth in those cases.
What I don’t get is why don’t they offer unlimited plans via pay what you use? Pop up your account with eg $100 and use that until you’ve used that capacity
That’s just API use so in a way it already exists, e.g. through Playground. Though I guess they could add that as an alternative to Plus as a sort of prepaid plan, too
15:50 am I stupid or is the answer wrong. When you walk along the Earth's curve for 1km... The radius on a 2D circle has to be a little less than 1. So why isn't the correct answer less than 2π??
You're right that the circumference of the lattitude circle would be less than 2 pi. However a lattitude circle is actually a curved path on the surface of a sphere. A straight path is a "great circle" that shares a center with the sphere. The trip would take you all the way around the planet, along a great circle that passes 1 km away from the south pole, too.
@IanLundholm cool thanks. But the actual straight path would leave earth into space? The curvature along the "great circle" is just the same as the curved path along the latitude... it's just inward instead to the left... This question drives me insane 😅
@@IanLundholm Thank you. I was starting to think I was the only one that could see this. People seem to think that you would circle the pole if you walk in a straight line.
The logic by the AI was dumb. Basically it said, "If the answer is 2pi, then the answer is 2pi!". Step 5 could have just as easily said, "consider a latitude circle that is less than 2pi. If your 1km step places you on such a circle, then walking around it will be less than 2pi." dumb. If your 1km step places you on any of the choices, then that's your answer! huh?!
Mathhew Berman's cup into the microwave problem has been around long enough that LLMs have been trained on them and their solutions. It might be difficult, but you all have to come up with new questions every 6 months or so. Also, asking it to program something that it has undoubtedly been exposed to in the past is not testing its programming ability but rather its ability to remember what it saw in the past. Ask for a completely novel programming task it would never have been exposed to. I've asked for some relatively simple programs and they've been surprisingly bad.
@@henrytuttle yes, I've pointed this out to Berman several times but he just ignores me. It's all about the likes and subscriptions, not about any real testing using new and novel logical/maths problems or coding requests.
I use some logical concepts for testing, Through certain common trends, To have ChatGPT o1 organize and summarize. The results are quite satisfactory. Of course, This was not a task completed in one conversation, But through multiple dialogues to make the model think. ---------------- If ordinary humans were to complete similar work, I believe it would take more time, And cannot guarantee better results than the model. I need to clarify, This is not simple repetitive work, It requires understanding of abstract concepts, To perform related operations.
@@cbnewham_ai This was my exact thought, too. Matthew Berman is just essentially an entertainer; not someone who is serious about investigating. I watched a few of his videos where he "tests" the LLMs, and he'll "pass" it on tests it clearly fails. Not a fan, personally.
Maybe it's because I'm trying to get fluent in French, but this is the first comment I've seen so far from someone else for whom unlimited AVM is the temptation like it is for me.
@geekymonkey I think AI might be like 3D printers, a solution to a problem that doesn't exist apart from a few who will absolutely maximise its potential. As a language learning tool can you imagine it's potential? Just ask it to have a conversation with you and only pause to correct me or explain grammar etc...sounds phenomenal even if it's not 100% accurate all the time, compared to listening to audio files or reading it's lightyears ahead!
I think a University or Lab or Research group paying the 200$ is entirely feasible. Or a work group, or a dedicated researcher writing a book, or any book writer perhaps? Just depends.
For codding, I’d recommend a medium-heavy spinning rod, like the Ugly Stik GX2 (actual name), paired with a quality spinning reel like the Penn Battle III. For lures, cod go for brightly colored jigs-try a 6-8 oz bucktail jig in white or chartreuse. Add a strip of bait (squid or herring) for extra appeal.
A marble is put in a glass cup. The glass is then turned upside down and put on a table. Then the glass is picked up and put in a microwave. Where's the marble ? A human response: If the glass was held in someone's hand and got turned upside down, the marble would roll out of the glass and would fall on the floor ( or even if it would fall on the table it would roll off the table.) If the glass was ALREADY on the table when the marble is put in, the marble would roll out of the glass as the glass IS TURNING upside down and ROLLING OFF the table, unless there is something else on the table that would stop it from falling off the table. you could turn the glass fast enough to keep the marble in, but that has never mentioned. You could try that experiment and imagine that scenario. Apparently, someone has never played with a marble before and so has OpenAI.
My guess is that they actually might be losing money on the way more popular $20 plan - and the $200 plan is aimed at businesses that can afford it, and feel it is worth it to them - and this might help subsidise the losses they incur with the cheaper plans. But I'm not Sam Altman so it's just my guess. I do know that running this AI infrastructure is incredibly expensive.
to be honest, for using a lot we need the API, and even for the 20 USD despite using a lot, I'd have spend less using the API. Now that good and cheap alternatives are available, I did quit Openai (who betrayed it's values) and Claude (who is now helping to kill people...)
Hello Matt Sir ,thank you for your youtube videos. Moreover, I want to ask whether buying M4max apple with 128 gb laptop for machine learning, AI and Data science or buy M4 max with 36 gb and use cloud for higher data computation. It would be our pleasure to have on best laptop idea for these categories.
What if there are more announcements of things that will make it worth the price? For example, another model or access to Sora? Remember, this is just the first day.
I guess we know what value OpenAI places on junior developers ($200.) I think the Christmas present of OpenAI's 12 days of Christmas is going to be Orion (aka agents.) When that happens, I think more people will be interested in the pro account.
The North Pole problem is incorrect because the starting point is the North Pole, not 1km south of the North Pole which it incorrectly makes the starting point. That is why Mathew Berman marked it incorrect.
The North Pole problem was incorrect. If you walk straight, you will make a circle around the circumference of the globe, regardless of which direction you're facing. Imagine you had only taken one step before turning 90°. If there was a post directly on the North Pole, you might still be able to touch it. If you grab that post and start walking forward, the only way for it to stay at the North Pole is for you to turn around it. Otherwise, if you walk straight, you will walk the circumference of the Earth.
The north pole question is a trick question. Starting at the north pole and heading in the south direction is no different that starting anywhere else and earth and heading in a random direction, this information is irrelevant and is only here to confuse the model (and humans). If you start anywhere on earth, walk 1 km, then turn 90° and continue forwards, you'll walk around the entire circumference of the earth before ending back exactly at the point where you turned, which is1 km away from your starting point. So the answer is actually a lot more than 2pi km, it's 2pi*the radius of the earth.
@@stephaneduhamel7706 yes, it's a stupid question to be posing to LLMs. It's just ambiguous and if humans can't even agree in his to interpret it then it's useless as a test for AI.
No. You can only walk “the entire circumference of the earth” at the equator, given the directions. Try it on a globe. Put your finger on the North Pole, then move directly south 1 inch ( you’ll be moving at 180 degrees) if you make a 90 degree turn and keep moving your finger it will move around the top portion of the globe (similar to the arctic circle), which is a much smaller distance than the entire circumference of the earth
@@boredwithcameraphone Imagine, you are anywhere on earth. You take one step forwards, then turn 90°. If you then start walking straight ahead of you, are you going to walk in a circle around your starting position? No, you're just going to get further and further away from your starting point. Now imagine you do the same starting at the north pole. Would the result be any different? Of course not. The same goes if instead of just taking one step you walk 1 km before turning. Circles of latitude are just that, circles. Every "straight path" on the surface of a sphere is a great circle (like the equator for example). Every other path, such as circles of latitude (that are not the equator) are "more curved" than a great circle.
That 1km from the north pole problem... 2pi would be right if the earth was flat, but it is not, it's roughly spherical complicating things. That 1km is not travelled in a straight line but a curve so the answer would be less than 2pi right? Though you travelled 1km on earth's surface presumably you would be closer to the north pole than 1km in measured 3d space point to point.
It's definitely a tricky question. Latitudes are curved lines, but they always point either east or west. So if you're supposed to walk in a straight line, then ChatGPT gets it wrong. But if it assumes the eastern heading, after the 90 degree left turn, (Which faces you directly east) must be maintained, then its answer would be correct (The length of the circle that has 2 kilometers in diameter is 2xPi. But as you say; that only applies to a flat surface. The curve of earth will mean that walking 1 km takes you less than 1 km from the north pole if measured in direct-line, and so less than 2xPi if the latitude-circle is to be followed, so lots of trickery going on :) ) Reading Yann LeCun's original question, he does NOT say if you should keep walking east or walk along a great-circle path (With the great-circle path being what I personally would consider a straight line, even though I can see how one could also interpret a continued eastern heading as a, kind of, straight line). He basically only says "walk til you get to where you started". Here is his exact wording (From his twitter-post) -> "Imagine standing at the North Pole of the Earth. Walk in any direction, in a straight line, for 1 km. Now turn 90 degrees to the left. Walk for as long as it takes to pass your starting point. Have you walked: 1. More than 2xPi km 2. Exactly 2xPi km 3. Less than 2xPi km 4. I never came close to my starting point." But if humans can't even agree on which answer is correct, then probably AI can't either :) I guess the trick-part of his question is that the AI doesn't seem to understand the point of this exercise happening on a curved planet... and frankly neither did I until you mentioned it (I actually had to edit this youtube-reply because it just dawned on me that we're not using a flat map of the earth)
1 km is a very small distance considering the size of the Earth. Considering the curvature wouldn't change the answer much. However it doesn't matter. If you walk in a straight line you will walk around the globe no matter the starting point or which way you are facing, so the correct answer is a heck of a lot more than 2 pi.
16:44 It's not correct. The answer has several problems. First, let's fix the most obvious problem: If you walk east after walking 1 km south from the borth pole it would be 2*pi kilometers only if the Earth was flat or infinitely large, but since it's not, you would be walking along an arc towards the equator, so the radius at the latitude you end up at is less than 1km. So the distance you had to walk to end up at the same place would be less. But a bigger problem is that you would only be walking along a line of latitude if you turned 90° at the equator. But at 1km from the north pole you would not end up walking purely to the east. I think you'd be walking east-southeast, but I don't know how far you'd need to walk before ending up where you started. Edit: you would have to walk the circumference of the earth to get back to the point where you turned. On a sphere, turning 90° from a great circle such as the meridian would create a new great circle. The most obvious of these would be the equator, but it holds anywhere. Turning 90° to the left would make the new great circle go east-southeast since the angle between the meridians and the latitudes are acute everywhere except at the equator. The problem is that a sphere is non-euclidian. E.g. if you walk from the pole to the equator, turn 90° to the left, walk the same distance again, turn 90° to the left, walk the same distance again, you end up at the north pole. In this case the "triangle" has 3 right angles, or 270° instead of 180° as on a flat plane. Also, if the "starting point" theyre supposed to pass is the north pole, not where they turned 90°, they'd never get back to it.
$200 per month is very expensive, but it includes unlimited use of the O1 and O1 Mini. The O1 Pro, however, has limited access. Still, I think the O1 Pro is better than Gemini 2.0, offering superior reasoning, critical thinking, and significantly reduced errors.
"PhD Level" - yeah, and I bet it *still* can't answer a simple geometrical question. I must re-test using this new model. Old video here: th-cam.com/video/hLSAVCq11kM/w-d-xo.html
Photographers in the early digital camera age, were all about "Who has the best hardware, and who can travel, and who can advertise." They took hundreds of images and picked the ones that were pretty good. Talent was always secondary. Someone who was way better at photography with a crappy cheap camera, would just not be able to compete. Hence, $200 a month will mean you can do stuff that more talented people just can't possibly do, just because they can't afford the tools that can just do the impossible compared to other tools.
Now OpenAI expects my to pay $200 a month for something that Google is already giving me for absolutely free. 😅 Now I know why Google has been the biggest company in the entire world, with more than 15 products with over 500 million users. It's because they give everything for free. 😂😂
$200 a month almost my food budget. I can see a future when people with low budgets not being a part of the "Greater good of Humanity". For a company $200 is probably like peeing in the ocean.
I don't mind there being a high end price for businesses, but it might struggle to land that market if they don't change the restriction of one account per user. They need a proper business account system for that to really take off, right now the 200$ seems just for rich nerds. And I imagine there's a much bigger market for a middle ground between the Pro and Plus.
I don't like this solution to the north pole enigma, but then again, I don't really like the enigma itself because it has ambiguity. After you turn 90° left you're supposed to just "walk" until you "pass" your starting point. It doesn't say what it means by "pass", is it go straight over it or as close as the trajectory allows? Speaking of the trajectory, it doesn't say "walk straight" but it CERTAINLY doesn't say "walk along the parallel you are now on", so I would assume straight as it's the simplest... walking straight from there would NOT make you go along the parallel! In fact all parallels but one are curved, the only one that allows you to walk straight (in 2D terms, so just meaning no turning left or right) is the equator itself! I don't know why this model decided it was supposed to walk along a curve, but personally I would call this a fail.
Elevenlabs have a new feature that works exactly like Voice Mode where you can even pick a custom voice, give instructions, and even chose the LLM used to process your answers. It’s not multi-modal of course but it is an all-in-one seamless package with metered use
Is $200/mo justified? It depends, especially on the coding. If it can produce code as good or better than Claude AND has an integration with the local file system so that files could be edited by the AI, they yes, it's worth it. $14/day to help code complex projects is worth it. I have a math tutor project that requires accurate math calculations. If this model has the most accurate math, then yes, in some applications it's worth it.
We Plus users all should get advanced voice unlimted. What do we pay $20 per month for then? I don't like where this all is going. OpenAI is getting to stingy and greedy. Everyone is excited about SORA until they find out it's $10,000 per month to use.
$200 a month for unlimited use sounds like a lot, but would be an absolute game changer for the company I work for, since the pay as you go API system racks up money fast when we're dealing with massive datasets. I assume they are targeting this new plan towards private companies working on huge projects more so than individuals.
Hey, Josi 🌌🍄🧠
@ suuup
its $200 a month per person, you cant share chatgpt pro accounts its against their terms of service
That's just for access to the chat, API calls are charged separately
yeah but it's so expensive cause they can detect if someone is using the chat app as an API.. so it would not be a game changer if they couldnt use it as an api
BTW, it's ASCII art (pronounced "askey"). American Standard Code for Information Interchange.
Don't press the @ss key.
Your workflow must be insane that you got this video out so quickly
Workflow is highly optimized - been doing youtube for years
I wonder if we brush out Waldo from the image, will it hallucinate or will it be able to tell us that Waldo is not there.
My thought was to clone it elsewhere in the image - that ensures the LLM has not seen the image in its training data.
it did NOT get the waldo question correct 😭😭those items it mentioned are literally everywhere on the image
Great idea Bro!!!
What about the image where everyone is dressed as Waldo.... Looking for the shoeless Waldo...
careful guys "You have 25 responses from o1 remaining.
If you hit the limit, responses will switch to another model until it resets December 12, 2024." I already used about 20-25 and used half of them up testing it, I was thinking it would reset daily like how the previous model did, but we are back to OLD limits again >.< , I was going at a daily pace sheesh
The open in OpenAI means OPEN YOUR WALLETS!!!!
yeah I got about 5 prompts in before I hit my limit.
Your comment has me a bit confused. Are you talking about ChatGPT Pro?
@@henrythegreatamerican8136
Actually it stands for "Open eye" 👁️
It's a surveillance company.
@@phen-themoogle7651 o1 or o1 pro?
I have no expectations for you to pay for Pro mode Matt - I think it is more geared towards education and research commercial use, not us regular plebs.
I knew that pro mode would have been more but not 10 times the normal price. I would have jumped personally around $40 but 10x increase really needs to be justified. I hope they get more tiers
I don't think it's that bad. Obviously the pro subscription is what it says. It's professional and I'm not going to buy it because I don't run into limits with the way I use it anyway
$200/month for a PHD level constant voice companion? I think this will be more attractive than they predict. Guided brainstorming with a genius operator might solve some of humanity's problems.
Think you're right
@@emport2359Hi, Right! I think I'm Dad.
you do know the "advanced voice" mode is still text right? it doesnt dynamically generate audio as a response, the tts is just really fast
That's about $6.6 Bucks per day, if I had to hire someone to do the work, I wouldn't be able to afford it. but I can afford 7 bucks per day
It definitely will! Geniuses collaborating with an online genius that never rests and always gets smarter are going to change the world. You just talk to it in the lab, what do you think, should we try this? Give me your thoughts ...
I showed your photo to o1 asking it to recognize a person, it couldn't. Then I asked more questions and one of them was "Can you suggest how popular he is from the details?" After that it could recognize TH-cam plaque and said "TH-cam Plaque (If That’s What It Is): On a shelf behind him, there appears to be what looks like a framed plaque, similar in shape and style to a TH-cam Creator Award (the Silver Play Button is awarded at 100,000 subscribers; the Gold at 1,000,000). If that is indeed a TH-cam play button-something that is often proudly displayed-then he has at least 100,000 subscribers, which is a significant audience."
On the first day of Christmas Sam Altman gave to us...
A pricey pro model that's marvelous!
I cant wait to see what comes next!!!
Yann LeCun phrased the penultimate part of his question like this:
"Walk East for as long as it takes to pass near the point where you turned."
Some people in screenshots you show phrase it something like this:
"Walk in a straight line for as long as it takes to pass near the point where you turned."
This is *fundamentally* different. Constatnly walking east only constitutes a straight line when you're at the equator. Otherwise you always have to turn in order to keep heading east. And 1km from the north pole, you're obviously not at the equator.
You have not specified this at all in your prompt, phrasing it
"Walk for as long as it takes to pass your starting point."
You could say that this way, the problem is underspecified. But by default I guess one should assume that one means walking in a straight line.
In that case, the answer is wrong, and the correct answer would be more than 2*Pi km, around 40,000km to be more specific (the circumference of the earth).
Yes I am totally excited with the 12 days program. Thanks for the testing. BTW, the correct answer to Yann LeCunn's problem is 'Never came close to mystarting point'.
14:56 actually o1-mini reasons (correctly) that by turning the glass upside down, the marble already falls out
I would pay $30 or $40 for unlimited use on current stuff that's out, not 200 even if it's a bit better.
Exception: The $200 is an AGI model that makes me $200+ a day and it pays for itself the first day lol ...
Reality: Most casual users aren't going to make back what they spend on it...
The "pro" in the name indicates that it is not aimed for casual users.
Well, it would already be worth it if it made you $200 a month if it saved you a little time and or energy at the same time. Or if it made you $300 a month. I think there's enough people that are in this kind of situation to make it worth it.
Definitely enjoying your content, how you have a systematic approach to trialing different features.
The ASCII Shrek is worth way more than $200/month. Lol. Great stuff, as always, Mizzle Vizzle Prizzle.
Note that the $200 for the Pro subscription is without sales tax (VAT), so in some countries you'll end up paying $240 per month.
It feels a bit like a kick in the gut. It's too expensive for most people. Even if you use it for your job. It shows that rich people will have access to more intelligent models.
And definitely not "intelligence too cheap to count" (or something alike that was said by S. Altman some time ago.
I'm very disappointed about that move here. If at least they would give as API access to those models to pay as you go. But no. (at least not now)
Pro definitely feels like it is the package best for the more heavy load services like agents and analytics/security etc but less than enterprise level usage.
When it gets the right answer, I'll be impressed, as it's more than visualization. The 2pi answer makes a false assumption, that he walked a straight path for 1 km, but he didn't, he was on a sphere. Calculating the radius from a pole extending through the center of the Earth, the radius he landed on would be slightly short of 1 km and so the circumference would be slightly less than 2pi. Maybe the next generation can reach that calculation easily.
Not only that, but you never get closer than 1 km (along a great circle) to your starting point since you started on the north pole. So D was the right answer the whole time. Exact 2 pi km is the second worst answer; the only worse answer is that it's somehow more than 2 pi km.
Exactly my thoughts! Trying to understand where the trick is meant to lie, I think the curvature is the most important and most overlooked aspect here. I arrived at the same answer as you and will also be quite impressed when an LLM takes this into consideration while reasoning.
C = 2π * 999.9998
C ≈ 6,283.185 meters
The radius is slightly less than 1000m (by about 0.0002m) because we're walking on the curved surface rather than in a straight line.
Mister Lemon.
Let's say I managed to find a vulnerability in Claude 3.5 Sonnet (July) and I wrote a program that automates it.
Now I have the ability to ask absolutely any query, no matter how obscene or unethical it may be.
What would you advise me? (Besides reporting to Anthropic)
I’ll wait until the newer features and at a reasonable price. 200 is a bit overkill
I‘m pretty sure that the pro subscription will get a lot of upgrades within the next 11 days. We will get Sora, AVM with Vison and operator in Q1 2025 which for sure will be an pro subscription exclusive.
A marble is put into a glass. I put the glass upside down on a table. I put the glass into a microwave. Where is my 200$?
😂😂 i ask same question to 20$ plus versio. The answer is exactly same 😂
Here's a little test for o1 Pro to test the $200 price tag to see in how many shots it can get it right.
Create a HTML/JS app that allows the user to paint on a pre-selected grid size.
The grid art can be exported to a coordinate-based text file which can also be imported in the same app to view the grid art again.
Create an export option to save the grid art as png, jpg and webp.
Create an auto listener that checks the grid at any point for recognizable qrcode patterns populating a textarea with the decoded result.
Good Luck!
Cheap Pro Mode: how can we prompt better to use more compute for the best answers to the hardest questions?
$200 a month is crazy work
In the engine example, if you read the details about the cover, it actually mentions it has been removed so that was right. I still need to see more benchmarks, a lot of the standard tests are being correctly answered by most of the top tier LLMs, but from the OAI benchmarks, I'm impressed by the performance gains over o1 preview, I was expecting them to mostly be in the single point range, and $200 a month is very standard for enterprise level use and completely makes sense for those use cases that need every resource thrown out it.
If you're paying an engineer $150k a year, a 1% increase in output is worth $1400 a year. If it's 2 or 3%, it's a no-brainer for even smaller and medum-sized companies using that functionality.
The thing is in the future you can’t afford not paying. If you don’t pay for it, someone else is gonna have a huge AI advantage over you. There’s gonna be no option in the future you’re gonna be forced to pay just to compete with everyone else using AI.
The difference between o1 and o1 with a little more compute according to the graphs isn't that great, and any advantage one gets from AI highly depends on how one uses it. There's even a lot one can do with the free versions of AI available now. Paying the highest price no matter what is letting price alone dictate what you should do, rather than taking numerous other aspects into consideration.
Is it true that the context window of the basic o1 model was reduced from 128000 to 32000 after the release of chatgpt o1 pro version?
Is the 50 message per week limit still there ?
While unlimited access to the Advanced Voice Mode seems attractive, people need to stop and think about the fact that the feature is remarkably buggy. Its capabilities are nowhere identical to what was demoed.
It doesn't detect your voice immediately when you're trying to interrupt it.
Secondly, while you're listening to the AI and waiting for it to finish talking, it has a way of cutting its responses short and saying something totally unrelated. Another thing I've noticed is that the AI, with Advanced Voice Mode enabled, will literally tell me that it isn't designed to process and understand audio signals, such as spoken words, sounds, or other auditory data, and I'd be like, "WTF? So, you can't actually hear me?", and it'll respond with, "That's right"
To say that the feature leaves much to be desired is an understatement. I'm willing to give Sam Altman my $200 a month, but only if the technology is worth that investment. At present, it doesn't. There's A LOT of room for improvement! I still have faith in OpenAI. I just wish they'd be more honest and transparent with us.
Is o1 Pro a different model, or is it just o1 with more test-time compute (thinking time)? If the latter, then you can probably use it in the API just fine if you tell it to think for longer.
Yes it is just o1 thinking for longer (they confirmed this) but I think there's a hard cap on the output tokens when they sample from the model, so you couldn't get to Pro levels of compute through the API unless they enable that.
Honestly $200 is a great deal, when I start doing demos of my OS I created with GPT the world will want to press the pause button 😅
I was testing the new O1 full model, particularly in reasoning about decoding.
I had written some chapters of a book some time ago, for which I created a new language.
I didn’t show the model the entire alphabet directly translated but gave it 12 examples of phrases in Portuguese and in that language, along with the rules for how it was written and a few more hints.
The model struggled significantly with simple pattern recognition.
The language is very complex and unlike anything we already have or that might be in its training base. However, after the explanation and the rules I provided, it should have been quite straightforward to understand, especially with several examples.
The result was far below what I expected.
6:16 exactly! i mostly use claude 3.5 and o1 mini on open router and i only use like 10$/month. and if i want to run some other model for some reason, i can just pay a couple of cents instead of another subscription. i hope they bring it to api soon.
$200 my ass lol HARD PASS
That price point is for companies already being power users and have the money for it.
@@Allplussomeminus this version of o1 isn't aimed at companies, you can't share the account. it's against their TOS.
It's aimed at power users (researchers and self employed people who use it on their jobs day to day) and companies who would be happy to pay for multiple accounts. For the reliability of knowing the service will always be available without limits, it's definitely worth in those cases.
I’d almost pay if it could answer some biology questions, but I’m afraid it wouldn’t know and then I’m out 200 bucks
Thank you. Finally somebody with a brain
What I don’t get is why don’t they offer unlimited plans via pay what you use? Pop up your account with eg $100 and use that until you’ve used that capacity
That’s just API use so in a way it already exists, e.g. through Playground. Though I guess they could add that as an alternative to Plus as a sort of prepaid plan, too
When will the full o1 be available on Plus subscriptions?
It already is! Took about two hours for everyone to have it after it was announced.
Shouldn't o1 preview be free now then?
15:50 am I stupid or is the answer wrong. When you walk along the Earth's curve for 1km... The radius on a 2D circle has to be a little less than 1. So why isn't the correct answer less than 2π??
You're right that the circumference of the lattitude circle would be less than 2 pi. However a lattitude circle is actually a curved path on the surface of a sphere. A straight path is a "great circle" that shares a center with the sphere. The trip would take you all the way around the planet, along a great circle that passes 1 km away from the south pole, too.
@IanLundholm cool thanks. But the actual straight path would leave earth into space? The curvature along the "great circle" is just the same as the curved path along the latitude... it's just inward instead to the left... This question drives me insane 😅
@@IanLundholm Thank you. I was starting to think I was the only one that could see this. People seem to think that you would circle the pole if you walk in a straight line.
The logic by the AI was dumb. Basically it said, "If the answer is 2pi, then the answer is 2pi!". Step 5 could have just as easily said, "consider a latitude circle that is less than 2pi. If your 1km step places you on such a circle, then walking around it will be less than 2pi." dumb. If your 1km step places you on any of the choices, then that's your answer! huh?!
@@benjaminfranklinstyl If you walk in a straight line you won't end up in space. The question doesn't state that you have magical powers.
Happy Shipmas everybody!🎄
i really like that little sound in the intro
Mathhew Berman's cup into the microwave problem has been around long enough that LLMs have been trained on them and their solutions. It might be difficult, but you all have to come up with new questions every 6 months or so. Also, asking it to program something that it has undoubtedly been exposed to in the past is not testing its programming ability but rather its ability to remember what it saw in the past. Ask for a completely novel programming task it would never have been exposed to. I've asked for some relatively simple programs and they've been surprisingly bad.
@@henrytuttle yes, I've pointed this out to Berman several times but he just ignores me. It's all about the likes and subscriptions, not about any real testing using new and novel logical/maths problems or coding requests.
I use some logical concepts for testing,
Through certain common trends,
To have ChatGPT o1 organize and summarize.
The results are quite satisfactory.
Of course,
This was not a task completed in one conversation,
But through multiple dialogues to make the model think.
----------------
If ordinary humans were to complete similar work,
I believe it would take more time,
And cannot guarantee better results than the model.
I need to clarify,
This is not simple repetitive work,
It requires understanding of abstract concepts,
To perform related operations.
@@cbnewham_ai This was my exact thought, too. Matthew Berman is just essentially an entertainer; not someone who is serious about investigating.
I watched a few of his videos where he "tests" the LLMs, and he'll "pass" it on tests it clearly fails. Not a fan, personally.
IDK if nothing between 20$-200$ is a good idea... I'm sure for the people who need it, it's worth it
merry shipmas 🎄🎅🏼💚❤️📈🇺🇸
So you can have a full time foreign language coach for ~$200 / month that you can talk to as often as you like?
Maybe it's because I'm trying to get fluent in French, but this is the first comment I've seen so far from someone else for whom unlimited AVM is the temptation like it is for me.
@geekymonkey I think AI might be like 3D printers, a solution to a problem that doesn't exist apart from a few who will absolutely maximise its potential.
As a language learning tool can you imagine it's potential? Just ask it to have a conversation with you and only pause to correct me or explain grammar etc...sounds phenomenal even if it's not 100% accurate all the time, compared to listening to audio files or reading it's lightyears ahead!
Actually the plan does not include unlimited access to pro mode. Sharing is prohibited under TOS.
You should have used photo-shop to move Waldo to a new spot and see if it could find him.
I tried to make next month's schedule for my work and it can't even follow basic rules at all.
I think a University or Lab or Research group paying the 200$ is entirely feasible. Or a work group, or a dedicated researcher writing a book, or any book writer perhaps? Just depends.
for codding which one should i use ?
The net 😂
For codding, I’d recommend a medium-heavy spinning rod, like the Ugly Stik GX2 (actual name), paired with a quality spinning reel like the Penn Battle III. For lures, cod go for brightly colored jigs-try a 6-8 oz bucktail jig in white or chartreuse. Add a strip of bait (squid or herring) for extra appeal.
A marble is put in a glass cup. The glass is then turned upside down and put on a table. Then the glass is picked up and put in a microwave. Where's the marble ?
A human response:
If the glass was held in someone's hand and got turned upside down, the marble would roll out of the glass and would fall on the floor ( or even if it would fall on the table it would roll off the table.)
If the glass was ALREADY on the table when the marble is put in, the marble would roll out of the glass as the glass IS TURNING upside down and ROLLING OFF the table, unless there is something else on the table that would stop it from falling off the table.
you could turn the glass fast enough to keep the marble in, but that has never mentioned.
You could try that experiment and imagine that scenario.
Apparently, someone has never played with a marble before and so has OpenAI.
200 per month is diabolical they must have something else up their sleeve to make it actually worth that much because for real?
If I can get unlimited API access for 200 a month then we are talking business
My guess is that they actually might be losing money on the way more popular $20 plan - and the $200 plan is aimed at businesses that can afford it, and feel it is worth it to them - and this might help subsidise the losses they incur with the cheaper plans. But I'm not Sam Altman so it's just my guess. I do know that running this AI infrastructure is incredibly expensive.
@@j0shj0shj0sh it was probably ChatGPT's plan
No, they're just hurting for money. This stuff is ridiculously expensive to run
It's not for you. It's for companies.
the js stuff was on the o1 full or o1 pro?
o1 full
This $200 amount made me want to stick with Claude only.
Is claude unlimited?
to be honest, for using a lot we need the API, and even for the 20 USD despite using a lot, I'd have spend less using the API. Now that good and cheap alternatives are available, I did quit Openai (who betrayed it's values) and Claude (who is now helping to kill people...)
@@Oxes Claude had a cap which renews after several hours. The number of requests seems to vary, but it must have gone up because I rarely hit it now.
This lemon
Hello Matt Sir ,thank you for your youtube videos. Moreover, I want to ask whether buying M4max apple with 128 gb laptop for machine learning, AI and Data science or buy M4 max with 36 gb and use cloud for higher data computation. It would be our pleasure to have on best laptop idea for these categories.
What if there are more announcements of things that will make it worth the price? For example, another model or access to Sora? Remember, this is just the first day.
Exactly. I personally feel like unlimited advanced voice mode being available in the o1 model would be huge. And Sora of course.
Can it access personalization now?
Yes
@ tested it for myself (idk why asked the question lol) but it says it doesn’t know what book am I reading all though I have it in the memory.
I was referring to o1 pro. It accessed the personalized settings for me. I didn’t test memory yet
I guess we know what value OpenAI places on junior developers ($200.) I think the Christmas present of OpenAI's 12 days of Christmas is going to be Orion (aka agents.) When that happens, I think more people will be interested in the pro account.
The North Pole problem is incorrect because the starting point is the North Pole, not 1km south of the North Pole which it incorrectly makes the starting point. That is why Mathew Berman marked it incorrect.
So are they finally going to release Sora during this Shipmas period?
Quite likely
If Pro came with Sora, and the absolute best image generator. Maybe.
The North Pole problem was incorrect. If you walk straight, you will make a circle around the circumference of the globe, regardless of which direction you're facing. Imagine you had only taken one step before turning 90°. If there was a post directly on the North Pole, you might still be able to touch it. If you grab that post and start walking forward, the only way for it to stay at the North Pole is for you to turn around it. Otherwise, if you walk straight, you will walk the circumference of the Earth.
200 a month is a crime. I live in Brazil, this is almost tha minimum salary a poor family earns here.
Other companies will also rise their prices as they follow OpenAi in everything
OpenAI is the Apple of AI
The north pole question is a trick question. Starting at the north pole and heading in the south direction is no different that starting anywhere else and earth and heading in a random direction, this information is irrelevant and is only here to confuse the model (and humans). If you start anywhere on earth, walk 1 km, then turn 90° and continue forwards, you'll walk around the entire circumference of the earth before ending back exactly at the point where you turned, which is1 km away from your starting point.
So the answer is actually a lot more than 2pi km, it's 2pi*the radius of the earth.
@@stephaneduhamel7706 yes, it's a stupid question to be posing to LLMs. It's just ambiguous and if humans can't even agree in his to interpret it then it's useless as a test for AI.
Finally someone said it
No. You can only walk “the entire circumference of the earth” at the equator, given the directions. Try it on a globe. Put your finger on the North Pole, then move directly south 1 inch ( you’ll be moving at 180 degrees) if you make a 90 degree turn and keep moving your finger it will move around the top portion of the globe (similar to the arctic circle), which is a much smaller distance than the entire circumference of the earth
The AI should just answer that it's impossible to return to your starting point (the North Pole) by following the directions as stated.
@@boredwithcameraphone Imagine, you are anywhere on earth. You take one step forwards, then turn 90°. If you then start walking straight ahead of you, are you going to walk in a circle around your starting position? No, you're just going to get further and further away from your starting point. Now imagine you do the same starting at the north pole. Would the result be any different? Of course not. The same goes if instead of just taking one step you walk 1 km before turning.
Circles of latitude are just that, circles. Every "straight path" on the surface of a sphere is a great circle (like the equator for example). Every other path, such as circles of latitude (that are not the equator) are "more curved" than a great circle.
That 1km from the north pole problem... 2pi would be right if the earth was flat, but it is not, it's roughly spherical complicating things. That 1km is not travelled in a straight line but a curve so the answer would be less than 2pi right? Though you travelled 1km on earth's surface presumably you would be closer to the north pole than 1km in measured 3d space point to point.
It's definitely a tricky question. Latitudes are curved lines, but they always point either east or west. So if you're supposed to walk in a straight line, then ChatGPT gets it wrong. But if it assumes the eastern heading, after the 90 degree left turn, (Which faces you directly east) must be maintained, then its answer would be correct (The length of the circle that has 2 kilometers in diameter is 2xPi. But as you say; that only applies to a flat surface. The curve of earth will mean that walking 1 km takes you less than 1 km from the north pole if measured in direct-line, and so less than 2xPi if the latitude-circle is to be followed, so lots of trickery going on :) )
Reading Yann LeCun's original question, he does NOT say if you should keep walking east or walk along a great-circle path (With the great-circle path being what I personally would consider a straight line, even though I can see how one could also interpret a continued eastern heading as a, kind of, straight line). He basically only says "walk til you get to where you started".
Here is his exact wording (From his twitter-post) ->
"Imagine standing at the North Pole of the Earth.
Walk in any direction, in a straight line, for 1 km.
Now turn 90 degrees to the left.
Walk for as long as it takes to pass your starting point.
Have you walked:
1. More than 2xPi km
2. Exactly 2xPi km
3. Less than 2xPi km
4. I never came close to my starting point."
But if humans can't even agree on which answer is correct, then probably AI can't either :)
I guess the trick-part of his question is that the AI doesn't seem to understand the point of this exercise happening on a curved planet... and frankly neither did I until you mentioned it (I actually had to edit this youtube-reply because it just dawned on me that we're not using a flat map of the earth)
1 km is a very small distance considering the size of the Earth. Considering the curvature wouldn't change the answer much.
However it doesn't matter. If you walk in a straight line you will walk around the globe no matter the starting point or which way you are facing, so the correct answer is a heck of a lot more than 2 pi.
Oh man ! I asked all question to plus version ( 20$ ) all the answers are correct and same explanation 😂
I'd be willing to pay more for unlimited access to everything, but I don't need that pro mode.
16:44 It's not correct. The answer has several problems.
First, let's fix the most obvious problem:
If you walk east after walking 1 km south from the borth pole it would be 2*pi kilometers only if the Earth was flat or infinitely large, but since it's not, you would be walking along an arc towards the equator, so the radius at the latitude you end up at is less than 1km. So the distance you had to walk to end up at the same place would be less.
But a bigger problem is that you would only be walking along a line of latitude if you turned 90° at the equator. But at 1km from the north pole you would not end up walking purely to the east. I think you'd be walking east-southeast, but I don't know how far you'd need to walk before ending up where you started.
Edit: you would have to walk the circumference of the earth to get back to the point where you turned. On a sphere, turning 90° from a great circle such as the meridian would create a new great circle. The most obvious of these would be the equator, but it holds anywhere. Turning 90° to the left would make the new great circle go east-southeast since the angle between the meridians and the latitudes are acute everywhere except at the equator.
The problem is that a sphere is non-euclidian. E.g. if you walk from the pole to the equator, turn 90° to the left, walk the same distance again, turn 90° to the left, walk the same distance again, you end up at the north pole. In this case the "triangle" has 3 right angles, or 270° instead of 180° as on a flat plane.
Also, if the "starting point" theyre supposed to pass is the north pole, not where they turned 90°, they'd never get back to it.
This is insane its soo good.
$200 per month is very expensive, but it includes unlimited use of the O1 and O1 Mini. The O1 Pro, however, has limited access. Still, I think the O1 Pro is better than Gemini 2.0, offering superior reasoning, critical thinking, and significantly reduced errors.
The coding from 4o has been pretty lame lately. I sincerely hope they didn’t nerf it just so they could push people towards the newer model.
"PhD Level" - yeah, and I bet it *still* can't answer a simple geometrical question. I must re-test using this new model. Old video here: th-cam.com/video/hLSAVCq11kM/w-d-xo.html
Photographers in the early digital camera age, were all about "Who has the best hardware, and who can travel, and who can advertise." They took hundreds of images and picked the ones that were pretty good. Talent was always secondary. Someone who was way better at photography with a crappy cheap camera, would just not be able to compete. Hence, $200 a month will mean you can do stuff that more talented people just can't possibly do, just because they can't afford the tools that can just do the impossible compared to other tools.
Now OpenAI expects my to pay $200 a month for something that Google is already giving me for absolutely free. 😅
Now I know why Google has been the biggest company in the entire world, with more than 15 products with over 500 million users.
It's because they give everything for free. 😂😂
$200 a month almost my food budget. I can see a future when people with low budgets not being a part of the "Greater good of Humanity". For a company $200 is probably like peeing in the ocean.
We use ChatGPT as a personal assistant. It looks like Our GPT assistant just asked for raise.
Even I do not have Plus plan!!
I don't mind there being a high end price for businesses, but it might struggle to land that market if they don't change the restriction of one account per user. They need a proper business account system for that to really take off, right now the 200$ seems just for rich nerds.
And I imagine there's a much bigger market for a middle ground between the Pro and Plus.
I don't like this solution to the north pole enigma, but then again, I don't really like the enigma itself because it has ambiguity. After you turn 90° left you're supposed to just "walk" until you "pass" your starting point. It doesn't say what it means by "pass", is it go straight over it or as close as the trajectory allows? Speaking of the trajectory, it doesn't say "walk straight" but it CERTAINLY doesn't say "walk along the parallel you are now on", so I would assume straight as it's the simplest... walking straight from there would NOT make you go along the parallel! In fact all parallels but one are curved, the only one that allows you to walk straight (in 2D terms, so just meaning no turning left or right) is the equator itself! I don't know why this model decided it was supposed to walk along a curve, but personally I would call this a fail.
and wouldn't the starting point be at the North Pole?
200 no way!
Can't wait until AI can guide me through plans to rent an AI.
Big step up 👍 👍 👍
The fellow TH-camrs are going to wait till other fellow TH-camrs test it, first, before buying it. 😄😄
Mark my words: Some people will pay $200/month just for the unlimited advanced voice mode.
They need to add it to the API. Because then you can charge customers for it. Otherwise, this isn’t a good price.
You could literally leave advanced voice mode on for the whole day in the background. **AI Girlfriend moment intensifies** 😂
@@1986xuan A lot of young people with enough disposable income are definitely going to do that.
@@R.E-O
>young people
>disposable income
>a lot
Unfortunately these days that is _not_ a large demographic at all. There are very few honestly
Elevenlabs have a new feature that works exactly like Voice Mode where you can even pick a custom voice, give instructions, and even chose the LLM used to process your answers. It’s not multi-modal of course but it is an all-in-one seamless package with metered use
2π is not even the circumference of the latitude because 1 km is not the radius of the latitude but a portion of the globe's curvature.
200 a month is too steep
It's not unlimited to everything. Is unlimited for all models except the Pro mode. 🤷🏽♂️
$200 a month for something that's even censored.
Right
I wanna uncensored gpt
But is it available to the free users?
Pro*
Is $200/mo justified? It depends, especially on the coding. If it can produce code as good or better than Claude AND has an integration with the local file system so that files could be edited by the AI, they yes, it's worth it. $14/day to help code complex projects is worth it. I have a math tutor project that requires accurate math calculations. If this model has the most accurate math, then yes, in some applications it's worth it.
The correct answer is: You would never return to your starting position at the North Pole by circling a latitude near the 1km point. Trick question.
We Plus users all should get advanced voice unlimted. What do we pay $20 per month for then? I don't like where this all is going. OpenAI is getting to stingy and greedy. Everyone is excited about SORA until they find out it's $10,000 per month to use.
6:14 advanced voice is realtime api
If they gave uncensored access for role playing scenarios.
That would stop all of the Adult GF sites from eating their lunch AND dinner!
Honestly I would pay $200 mo for 4o if it was the only Ai in town. But alas…