Regarding the DeepSeek topic: Why shouldn't this be a big deal? OpenAI's claim of theft is fundamentally flawed. They argue DeepSeek "stole" something, but let's be clear - if anything was "stolen" was already circulating. Their primary accusation involves using OpenAI outputs, but here's the critical context: we're talking about a reasoning model with a sophisticated token context that OpenAI never publicly disclosed and doesn't make accessible through their API. DeepSeek has made significant technical breakthroughs that deserve acknowledgment, not dismissal through corporate spin. The attempts to downplay their achievements seem entirely disingenuous. The fundamental fact remains: they've overtaken OpenAI through their own innovative efforts. Their models aren't just derivative - there are also some foundational differences that go back to the most basic computational levels like them not using standard CUDA. The real story here is the numerous technical improvements DeepSeek has introduced and potential for unleashed reinforcement learning. That's the big news - and it's being obscured by petty corporate rhetoric.
100% agree, is funny to see "Open"AI's response with claiming that its all stolen when most of the data they have taken and used for their training is stolen themselves, and the fact that DeepSeek R1 is so much more open and that their model is actually downloadable and you can run it on your own machine creates a stark contrast to OpenAI's actual openness and agenda
Yep you're correct and also Sam Altman is such a sore loser, he is so full of himself he can't even comprehend that some really smart people at the other side of the world can make huge advancements in AI other than him xd
Saying DeepSeek is "not that impressive" because its training process involves using existing models, is like saying the invention of Ford Model T is not impressive because it borrowed ideas from the Motorwagen.
It's not impressive because the numbers are fake. Actual development costs are like $5B. And number of Nvidia cards used are closer to 50,000! reason they under reported it is to cause shock and awe and also they don't want to reveal they got more GPU than allowed under sanction
@Benny-tx5qd Deepseek wins this round. When queried about what model it is. Deepseek actually returns that it's ChatGPT. They can't even copy homework! In high school, the first rule for copying homework is, change the name of your classmate to your own name!
Deepseek can not be banned because a ban is not enforceable on an open source backend service. Devs already have copies. If they decide to run Deepseek on a private server no one knows their frontend app is making requests to it. TikTok is a frontend app that depends on app stores for access and therefore can be banned.
DeepSeek exposed the hyperinflated AI market that has for so long been dictated by the US companies... I have downloaded and I am now running my own model based on DeepSeek and I have never been more excited for the future.
@@Jst4vdeos running it locally you can litterally sandbox it without any internet access or ways to get the data away from your own devices. i would be more critical to sharing data to their or OpenAI's services directly where you have no control over it
@@Jst4vdeos he is running it local so there is no way anything is leaked by just talking to it. U can remove limits too. And literally talk about anything.
Deepseek feels like the moment when Apache was released in the 90s. Netscape used to charge $1500-5000 for a web server and had 60% market share in 1995. Then Apache was released for free and Netscape folded up its Web Server department within 4 years.
DeepSeek is not at all about WHAT IT IS TRAINED ON (internet data or ChatGPT outputs). But more about HOW IT IS TRAINED, which makes it better as well as more efficient. OpenAI's claims are stupid, as even the Gemini has some training data from Output generated by ChatGPT. OpenAI does not give access to their models to anyone, so DeepSeek being "TRAINED ON" their model is just not true. edit: also just cause some training data was OUTPUT from ChatGPT doesn't mean it is stolen. Especially from a company which itself "STOLE" data to train the ChatGPT in the first place. and if so, then Gemini, and many other LLMs are also doing the same. It is not about Training Data, but Training Method which is why DeepSeek became talk of the town. And OpenAI is just trying their hard to coverup their a$$e$ for wasting so much of investor's money
From what I understand they stole the parameters that chatgpt input into their model, like they literally somehow copied the whole data chatgpt put the effort to collect, and used the same data to train their model. This has to be illegal because that is confidential information, even if it is public data no one should have the exact copy of that data. So basically they somehow stole it from the private datacenter of chatgpt.
Nope, I have some experience from making LLMs, large language models are completely based on big amounts of data and vectors connecting key words. So model is basically calculating what is statistically most likely correct answer. So data is the most important part, also verifying the veracity of the data. Model can be also trained to learn on past mistakes, but it is hard to verify what data is correct. Also, some companies make millions just by selling data for learning LLMs. But, I think it is hard to prove that they were stealing data because they weren't taking it from the source, but the publicly accessible results of ChatGPT.
@@musicb5202they paid to query ChatGPT and llama . OpenAI scraped data from the internet without regard for copyright or intellectual property ingested it into a database organised transformed and used it to train its model then sells service based on that LLM and demand that user cannot train their models . It like I woke up and realised that most commentators have a bias that impacts their reporting on science I feel so disappointed it’s shocking the level of computer illiteracy is astounding
It gets even worse for the big US AI corporations, Jiayi Pan, a PhD candidate at the University of California, Berkeley, claims that he and his AI research team have recreated core functions of DeepSeek's R1-Zero for just $30 - a comically more limited budget than DeepSeek, which rattled the tech industry this week with its extremely thrifty model that it says cost just a few million to train. It appears that the DeepSeek model is exceptionally fast at learning compared to other models making it so cheap to train and efficient then add to that the fact it's so cheap to run. Good news for all but the big US AI corporations.
is great deekseek have open sourced vvestern soft power given deepseek relies on a parameterized westerm cultural corpus to do the r1 model's heavy lifting by extracting implicit liberal individual western chain of thought reasoning norms which the illiberal chinese corpus is largely bereft of. *me asking deepseek:* "given english is the lingua franca of many domains then surely you bias toward english chain of thought?" *deepseek's reply:* "You're absolutely right to point out that English, as a global lingua franca, plays a significant role in shaping my training data and, by extension, my responses. This does introduce a bias toward English-based patterns of thought, especially in domains where English dominates, such as science, technology, and international discourse."
Using a fraction of the power and ability to run on minimal hardware to do the same job is a REVOLUTION that Deepseek will be know for. Releaseing this amazing saving of power to the entire world cannot be understated.. Even if you want to ;-)
The Concorde had a maximum cruising speed of approximately 1,354 miles per hour (2,179 kilometers per hour), which is more than twice the speed of sound, or Mach 2.04. This speed allowed it to fly from New York to London in as little as 2 hours, 52 minutes, and 59 seconds, setting a record for the fastest transatlantic crossing by a passenger aircraft. To achieve such high speeds, the Concorde was designed with features like a streamlined body, swept-back wings, and powerful engines equipped with afterburners, which provided additional thrust and efficiency.
I watch/listen to the podcast when every episode releases while I work. Best thing to listen to when you’re a cable guy replacing wires up top a pole or on the roof of a house 😂👌🏾
Fun fact, my dad used to take the Concorde for work trips from the US to Europe! He said it was crazy to see the earth from that perspective. Also it was extremely loud during the flight
This has been one of my favorite episodes! I would love to see just David and Marques more often. David gives these amazing deep-dives on interesting topics, and Marques responds with his expert opinions. I hate to say this, but Andrew often takes away from the conversation.
Day 1 of asking for the following things: 1. Having Andrew dress up as Bobby Slowik 2. Implementing a rule where if a member wants to make any prediction they must wager a certain amount of trivia points.
IMO, what people are missing with the DeepSeek and stock panic, is that just because DeepSeek figured out how to do more with less powerful and less quantity of GPUs and what makes people think that the next iteration won't use the DeelSeek methodology, but with all the powerful GPUs? More is more is more. Also, I think the idea the DeepSeek "stole" training data from OpenAI is cope from OpenAI, to minimize what DeepSeek has accomplished. Finally, these closed source AI companies (OpenAI, Anthropic, Google, etc. etc.) have all been touting the coming of S-AGI, where mankind can allow the AIs and the robots do the work, but they don't address the fact that they would own the AIs and the robots. They are essentially describing a nightmare future where we are all reduced to vermin, scurrying around for survival, because we would completely rely on the 10 companies in the world that owned the means of creation, production, transportation, etc. etc. etc. The only way we ALL get to the "Bright" future that they are touting, the Star Trek future, is if the AIs are open source and available to all.
@@xuezhongyu5853 Xaclty!! Also, in this situation, Prometheus is Chinese and The United States of Olympus has been making China the boogie man for decades. If only the people of Olympus realized that capitalism is their enemy and a capital owned AI will be their destruction.
Creation, production, transportation, etc. for who? Its for us, why would they waste their resources to efficiently produce and distribute product if they didnt care? If they dont want to waste resources and just stopped trying to produce and all that stuff, then its ours for the taking. You also don’t consider how much would happen before a nightmare reality exists. If AI gets intelligent enough, it will realize whether its owners are evil or not.
@@jaydennguyen-xk1yo It is for our purchasing, but not for our profit. Capital owns the AI and essentially everything. No one will be able to unseat their power, because nothing could achieve anything to compete with them and their AI. This is the time for "how much would happen before a nightmare reality exists" and nothing is happening. Finally, I am not sure it is a viable plan to depend upon AIs being "evil or not". Why would the AI's owner their AI to be anything but self serving to themselves?
@@sevilnatas Because AI will either be as good as experts or better in a nightmare scenario. If they are as good as experts, then the experts can replicate their knowledge. If they are better, then nobody in the world could influence what the AI knows because it creates its own knowledge and its own reasoning, the owner wont even have control because its above the owner’s intelligence. The people making all the stuff and doing all the stuff depend on us, we dont have to buy from them, and how would we buy from them if AI is doing all the jobs? Imagine if you were an AI training itself on its own information, you influence yourself and the path forward is for AI to have bodies to interact with the world and learn from the data that we learned from. Its just a human that can learn more and at a much faster pace. Would you bet on a super-intelligent human being evil or good? If you think good is rational and evil is irrational then its obvious what the answer would be. Sure an evil person could come along and try to make malicious AI but thats also a threat with the open source AI and almost anything in the world. It also takes a tremendous amount of work to be able to manage the entire supply chain of all these products and the logistics. If AI is doing for the owners then the owner depends on the AI. I still think many things would happen before the nightmare reality. The billions on this earth wouldnt just slowly let companies swallow whole industries
31:59 As someone who is familiar with Samsung's policies, I believe the initial price of the handset will be $2,200-$2,000, and then the price will drop after a couple of months until the average consumer thinks it's a bargain.
1:28:36 I believe Boom Overture will be designed to divert some of the sound energy created by a sonic boom upwards so as to reduce the noise levels of supersonic flight above land.
Yes, and Concorde did NOT fly supersonic over land, per regulations to lower sound pollution. There was no NY to CA routes for example. Which limited its usefulness, and was one of the reasons they only made about 20 of them instead of the hundreds initially projected to be sold. Anyone wanting to know more about Concorde should read Mike Bannister’s book, called Concorde. It’s an amazing aircraft.
BTW if you remember the M.U.G.E.N engine, which involved in lots of 2D fighting games, its name in Japanese (“無限”) means the same as Moohan in Korean (“무한”). Both name derived from the Chinese word “無限”(wú xiàn), yes “infinity”.
DeekSeek's inference cost is really cheap, but it isn't clear what the actual cost of retraining or updating the model is. The localized versions of DeekSeek are basically a snap-shot of the available training data that doesn't have new information.
I am so excited for Pebble coming back. I was a kid when it was first around and always wanted one. e-ink just makes so much sense for a watch, I don't need or want to be able to scroll Instagram or whatever. The Pebble just got the basic needs done right and done perfectly.
15:44 This demo of Google maps in the virtual world looks so awesome, vivid and with rich detailed simulation. I can't wait to see when Google integrates this type of simulation to make it easy for people to find their way around places. Just imagine for you to get a clear popup image of your desired location within seconds of inquiring on your phone. That will be really super awesome 😎💯💪🏾👍🏾
53:00 Not to brag but my Garmin smartwatch has an AMOLED display, does fitness tracking, and all of the notification stuff while lasting for 5 days (27 days if I put it in power saving mode). Garmin also sells some more hardcore models with solar charging that you literally never have to plug in
58:45 You mean like 90% of sub $50 smartbands and watches? Amazfit have a whole set of those. They even have some basic heart rate sensors, but you can disable or ignore that. It' just shows the time - and you can customize the "watchface" completely. Show everything. Or keep it minimal, like me. I have just red retro digital numbers on a black background for time. And the screen pops up when I get a notification. The battery lasts a week or so. E-inkt versions last even longer.
These high speed planes are literally being promoted and funded by Saudi Arabia because they consume so much jet fuel. How can we even talk about investing high carbon technologies in the current state of the world. Honestly hope these planes never get used
Hell yes I would go on a free Boom flight! One of my few regrets in life is putting off flying on the Concorde. It’s amazing. I encourage you all to go on the Concorde tour at the Intrepid Museum; seeing it in person really shows how different it is compared to regular airliners.
If you just want a smart watch with basic functions on the cheap, why not just get a Huawei watch? 2 weeks of battery life, HR/O2sat/step tracking, notifications, music control, big screen, and pretty stylish. If you have android you can even sideload the international version of the app and activate more advanced functions like GPS, app store, watch face store, and if you have the Pro version (more expensive though) then stuff like BP monitoring and EKG as well.
MKB Awesome Podcast.Welcome back.....Love the shirt also thank You guys for sharing the FurBaby on the Team.I also miss the Pebble Watch especially with the leather band.Thank You for sharing all updates.💯✌️
I'm still using my Pebble Time Steel and Pebble Time. Nice to know about my notifications or calls I can ignore with a glance without having to take out my phone or keep it on me at home or office. Always using it for media control and while I've cracked the glass from dropping it on cement, I've stopped even checking after scrapping against a wall because it's always buffed right out. Except for the Round they were all water resistant while Apple's first watch like Pebble's Round was only splash resistant. I'd still be using my Pebble Steel for it's style and higher contrast if I'd replace the zebra-stripe connector. I have not replace the batteries and they still last several days thanks to the Time Steel's originally 10 day battery.
I love how the video playing inside of android xr at 10:15 is the exact same video Ellis talks about in one of the previous episodes he watches while on facetime
I'd love a new Pebble watch. I went from Fossil to Samsung to Fitbit and now use my Casio Illuminator. If this thing buzzed and had a notification LED, it would be perfect. Also, put me on the first flight.
Is the podcast no longer posting in audio-only format anymore? The last 2 episodes defaulted me to the video format and worries me about data consumption.
I had a flight once that was delayed by a half hour before we heard over the intercom "We sincerely apologize, preflight checks revealed an issue with the aircraft so the pilot has made the decision to take a plane from the boneyard to ensure your safety. Please stay patient as we take another hour to ensure the plane is ready for flight"
So many of our planes are already edging the speed of sound -- most of my flights recently are in the 750 and 760mph range, the most I've seen is 767mph when we were in a jetstream! None of these create sonic booms because they're being pushed by the air around them, so relative to that air it's still slower.
@44:47 The paradox you're referring to is known as Jevons Paradox. It states that as technological improvements increase the efficiency of resource use (or as the cost of a product decreases), overall consumption of that resource (or product) can increase rather than decrease.
Regarding the Pebble, Quinn from Snazzy Labs just reviewed an e-ink tablet, TRMNL, that costs $120, available now, which can display essentially anything from your phone/computer, months-long battery life. Not something I needed but ordered one this morning anyway; hope you guys can review it soon. EDIT: oh, and flying in the jetstream does NOT burn more fuel; it just isn't there all the time.
People really forget Googles business model. Let’s give this google operating system access to where I’m looking whenever I’m wearing it, also where I’m not looking and with microphones. I bought and returned my Quest 2 for this same reason.
Ellis got stuck in Apple Vision Pro, that's why he's missing this Friday
Oh my, he was right all along. A magical piece of trash indeed.
hopefully his step-sibling can help him out 🙏😔
I have an idea for the next video: first person view! 😂
Regarding the DeepSeek topic: Why shouldn't this be a big deal? OpenAI's claim of theft is fundamentally flawed. They argue DeepSeek "stole" something, but let's be clear - if anything was "stolen" was already circulating. Their primary accusation involves using OpenAI outputs, but here's the critical context: we're talking about a reasoning model with a sophisticated token context that OpenAI never publicly disclosed and doesn't make accessible through their API. DeepSeek has made significant technical breakthroughs that deserve acknowledgment, not dismissal through corporate spin. The attempts to downplay their achievements seem entirely disingenuous. The fundamental fact remains: they've overtaken OpenAI through their own innovative efforts. Their models aren't just derivative - there are also some foundational differences that go back to the most basic computational levels like them not using standard CUDA. The real story here is the numerous technical improvements DeepSeek has introduced and potential for unleashed reinforcement learning. That's the big news - and it's being obscured by petty corporate rhetoric.
Nope
@@dan-kt6vm Wow, the detailed argumentation blows me away! Now, jokes aside, with what do you disagree? With everything?
100% agree, is funny to see "Open"AI's response with claiming that its all stolen when most of the data they have taken and used for their training is stolen themselves, and the fact that DeepSeek R1 is so much more open and that their model is actually downloadable and you can run it on your own machine creates a stark contrast to OpenAI's actual openness and agenda
Yep you're correct and also Sam Altman is such a sore loser, he is so full of himself he can't even comprehend that some really smart people at the other side of the world can make huge advancements in AI other than him xd
The accusations just scream sore loser mentality. Prolly has a lot to do with the amount of people unsubbing from ChatGPT.
The irony of OpenAI crying about others stealing their stolen content. Davids reaction is totally valid 😂
When big tech steals from us it’s fair use but when someone steals from them we need to ban it 😂
Next
OPEN?AI steal your data: for great good
DS steal your date: nono
Another day, another classic double standard
Everyday life in America
Ikr 😂😂
Gemini自曝中文用百度文心一言训练
Gemini revealed that it used Baidu Wenxin Yiyuan for Chinese language training
Saying DeepSeek is "not that impressive" because its training process involves using existing models, is like saying the invention of Ford Model T is not impressive because it borrowed ideas from the Motorwagen.
Wait till he hears about Qwen 2.5
It's not impressive because the numbers are fake. Actual development costs are like $5B. And number of Nvidia cards used are closer to 50,000! reason they under reported it is to cause shock and awe and also they don't want to reveal they got more GPU than allowed under sanction
Gemini自曝中文用百度文心一言训练
Gemini revealed that it used Baidu Wenxin Yiyuan for Chinese language training
@Benny-tx5qd Deepseek wins this round. When queried about what model it is. Deepseek actually returns that it's ChatGPT. They can't even copy homework!
In high school, the first rule for copying homework is, change the name of your classmate to your own name!
It doesn't say it's based upon ChatGPT, it says it's DeepSeek V3 developed by DeepSeek Inc., and not Chat GPT
Deepseek can not be banned because a ban is not enforceable on an open source backend service. Devs already have copies. If they decide to run Deepseek on a private server no one knows their frontend app is making requests to it. TikTok is a frontend app that depends on app stores for access and therefore can be banned.
in most parts of the world. Third-party app stores are openly possible, but only in individual places have Google and Apple monopolies.
The deepseek ban is more about the app and API, not the locally run model.
@@user-sf8du It's whack-a-mole because new apps would spin up quickly.
There is no issue with the model itself just the servers in china it’s being run on
Deepseek is the best non-boring thing that has happened in the battle for supremacy in AI world and I f*cking love it.
DeepSeek exposed the hyperinflated AI market that has for so long been dictated by the US companies...
I have downloaded and I am now running my own model based on DeepSeek and I have never been more excited for the future.
Careful what personal data you expose to it
@@Jst4vdeos running it locally you can litterally sandbox it without any internet access or ways to get the data away from your own devices.
i would be more critical to sharing data to their or OpenAI's services directly where you have no control over it
@@Jst4vdeos bro it runs local you noob.
@@Jst4vdeos he is running it local so there is no way anything is leaked by just talking to it. U can remove limits too. And literally talk about anything.
where can you download it? 🥺
DeepSeek research papers are impressive
Yes! Too bad politics is rearing its ugly head. AI knows no borders. It's for everyone!
Deepseek feels like the moment when Apache was released in the 90s. Netscape used to charge $1500-5000 for a web server and had 60% market share in 1995. Then Apache was released for free and Netscape folded up its Web Server department within 4 years.
Thanks you for sharing this, I haven't heard this story before
ClosedAI “But it stole our outputs!”
*Musicians, writers, programmers and artists playing the worlds smallest violin*
DeepSeek is not at all about WHAT IT IS TRAINED ON (internet data or ChatGPT outputs).
But more about HOW IT IS TRAINED, which makes it better as well as more efficient.
OpenAI's claims are stupid, as even the Gemini has some training data from Output generated by ChatGPT.
OpenAI does not give access to their models to anyone, so DeepSeek being "TRAINED ON" their model is just not true.
edit: also just cause some training data was OUTPUT from ChatGPT doesn't mean it is stolen.
Especially from a company which itself "STOLE" data to train the ChatGPT in the first place.
and if so, then Gemini, and many other LLMs are also doing the same.
It is not about Training Data, but Training Method which is why DeepSeek became talk of the town.
And OpenAI is just trying their hard to coverup their a$$e$ for wasting so much of investor's money
From what I understand they stole the parameters that chatgpt input into their model, like they literally somehow copied the whole data chatgpt put the effort to collect, and used the same data to train their model. This has to be illegal because that is confidential information, even if it is public data no one should have the exact copy of that data. So basically they somehow stole it from the private datacenter of chatgpt.
Nope, I have some experience from making LLMs, large language models are completely based on big amounts of data and vectors connecting key words. So model is basically calculating what is statistically most likely correct answer.
So data is the most important part, also verifying the veracity of the data. Model can be also trained to learn on past mistakes, but it is hard to verify what data is correct. Also, some companies make millions just by selling data for learning LLMs.
But, I think it is hard to prove that they were stealing data because they weren't taking it from the source, but the publicly accessible results of ChatGPT.
@@musicb5202they paid to query ChatGPT and llama . OpenAI scraped data from the internet without regard for copyright or intellectual property ingested it into a database organised transformed and used it to train its model then sells service based on that LLM and demand that user cannot train their models . It like I woke up and realised that most commentators have a bias that impacts their reporting on science I feel so disappointed it’s shocking the level of computer illiteracy is astounding
@@leonardocrncec9566If they created mass requests to the API, the host company would notice and put a stop to it.
35:31 “they stole our stolen content” should become the newest viral meme about ai-anything. With his crying gesture and all. 😅
Would be annoying
Use the new feature from the Galaxy 25😊
😂😂
Gemini自曝中文用百度文心一言训练
Gemini revealed that it used Baidu Wenxin Yiyuan for Chinese language training
44:45 It’s called “Jevons Paradox” for those who are curious.
Nobody is curious buddy
@@MrMikeymontemayorDon't assume others are as incurious as you are
@@MrMikeymontemayor I am, i specifically used word finder for "paradox" to arrive to this comment. Thanks to OP. I bet you're fun at parties Mikey.
It gets even worse for the big US AI corporations, Jiayi Pan, a PhD candidate at the University of California, Berkeley, claims that he and his AI research team have recreated core functions of DeepSeek's R1-Zero for just $30 - a comically more limited budget than DeepSeek, which rattled the tech industry this week with its extremely thrifty model that it says cost just a few million to train. It appears that the DeepSeek model is exceptionally fast at learning compared to other models making it so cheap to train and efficient then add to that the fact it's so cheap to run.
Good news for all but the big US AI corporations.
lol, truly Chinese Americans vs Chinese Chinese.
I am so happy China is doing this!! I love DeepSeek ❤❤❤❤
We need more competition
Deepseek is the best non-boring thing that has happened in the battle for supremacy in AI world and I f*cking love it.
is great deekseek have open sourced vvestern soft power given deepseek relies on a parameterized westerm cultural corpus to do the r1 model's heavy lifting by extracting implicit liberal individual western chain of thought reasoning norms which the illiberal chinese corpus is largely bereft of.
*me asking deepseek:* "given english is the lingua franca of many domains then surely you bias toward english chain of thought?"
*deepseek's reply:* "You're absolutely right to point out that English, as a global lingua franca, plays a significant role in shaping my training data and, by extension, my responses. This does introduce a bias toward English-based patterns of thought, especially in domains where English dominates, such as science, technology, and international discourse."
@@blengi interesting
@@blengi substitute all "western" with "modern" in your argument would make it much more accurate.
2 pods in a week♥ Welcome back Marques.
$699 or $799 is straight up DELUSIONAL pricing.
🤣🤣🤣
Using a fraction of the power and ability to run on minimal hardware to do the same job is a REVOLUTION that Deepseek will be know for. Releaseing this amazing saving of power to the entire world cannot be understated..
Even if you want to ;-)
During Android XR segment, DAVID you asked the perfect questions.
Deepseek actually helps the AI industry by showing there is a different way to work this game out.
The Concorde had a maximum cruising speed of approximately 1,354 miles per hour (2,179 kilometers per hour), which is more than twice the speed of sound, or Mach 2.04. This speed allowed it to fly from New York to London in as little as 2 hours, 52 minutes, and 59 seconds, setting a record for the fastest transatlantic crossing by a passenger aircraft. To achieve such high speeds, the Concorde was designed with features like a streamlined body, swept-back wings, and powerful engines equipped with afterburners, which provided additional thrust and efficiency.
This podcast is the favorite part of my Friday mornings. Thanks to all who make Waveform possible!
So this is how I find out that 5pm for me can be 11am for others.
@@just_mdd4 Grandma used to say 'It's 5 O'clock somewhere' as she made herself a drink at any time of the day.
I watch/listen to the podcast when every episode releases while I work. Best thing to listen to when you’re a cable guy replacing wires up top a pole or on the roof of a house 😂👌🏾
The funniest take on Deepseek I read somewhere: chatGPT lost its job to AI
DISHES TIME LADS
You can run a local version of deepseek on a Mac mini M1. It’s quite cool to use
Fun fact, my dad used to take the Concorde for work trips from the US to Europe! He said it was crazy to see the earth from that perspective. Also it was extremely loud during the flight
57:16 You are describing Garmin forerunner watches amazing battery, simple watch face , super clear in the sun and affordable
We love finally seeing a competitive market of A.I
The customers will get the most benefits the more competitor are in the market 🎉🎉🎉
Deepseek did shook the AI world big time
This has been one of my favorite episodes! I would love to see just David and Marques more often. David gives these amazing deep-dives on interesting topics, and Marques responds with his expert opinions. I hate to say this, but Andrew often takes away from the conversation.
Day 1 of asking for the following things:
1. Having Andrew dress up as Bobby Slowik
2. Implementing a rule where if a member wants to make any prediction they must wager a certain amount of trivia points.
1:04:33 - david lil smirk, he is so pround🤌
Damn the bonus episode really messed up my clock.
Hope we get back to a full cast episode so we can get Trivia again😅
There's trivia on this ep!
What are you talking about?
IMO, what people are missing with the DeepSeek and stock panic, is that just because DeepSeek figured out how to do more with less powerful and less quantity of GPUs and what makes people think that the next iteration won't use the DeelSeek methodology, but with all the powerful GPUs? More is more is more. Also, I think the idea the DeepSeek "stole" training data from OpenAI is cope from OpenAI, to minimize what DeepSeek has accomplished. Finally, these closed source AI companies (OpenAI, Anthropic, Google, etc. etc.) have all been touting the coming of S-AGI, where mankind can allow the AIs and the robots do the work, but they don't address the fact that they would own the AIs and the robots. They are essentially describing a nightmare future where we are all reduced to vermin, scurrying around for survival, because we would completely rely on the 10 companies in the world that owned the means of creation, production, transportation, etc. etc. etc. The only way we ALL get to the "Bright" future that they are touting, the Star Trek future, is if the AIs are open source and available to all.
Prometheus unleashed fire, now everyone can use "fire".So Olympus is angry now.
@@xuezhongyu5853 Xaclty!! Also, in this situation, Prometheus is Chinese and The United States of Olympus has been making China the boogie man for decades. If only the people of Olympus realized that capitalism is their enemy and a capital owned AI will be their destruction.
Creation, production, transportation, etc. for who? Its for us, why would they waste their resources to efficiently produce and distribute product if they didnt care? If they dont want to waste resources and just stopped trying to produce and all that stuff, then its ours for the taking. You also don’t consider how much would happen before a nightmare reality exists. If AI gets intelligent enough, it will realize whether its owners are evil or not.
@@jaydennguyen-xk1yo It is for our purchasing, but not for our profit. Capital owns the AI and essentially everything. No one will be able to unseat their power, because nothing could achieve anything to compete with them and their AI. This is the time for "how much would happen before a nightmare reality exists" and nothing is happening. Finally, I am not sure it is a viable plan to depend upon AIs being "evil or not". Why would the AI's owner their AI to be anything but self serving to themselves?
@@sevilnatas Because AI will either be as good as experts or better in a nightmare scenario. If they are as good as experts, then the experts can replicate their knowledge. If they are better, then nobody in the world could influence what the AI knows because it creates its own knowledge and its own reasoning, the owner wont even have control because its above the owner’s intelligence. The people making all the stuff and doing all the stuff depend on us, we dont have to buy from them, and how would we buy from them if AI is doing all the jobs? Imagine if you were an AI training itself on its own information, you influence yourself and the path forward is for AI to have bodies to interact with the world and learn from the data that we learned from. Its just a human that can learn more and at a much faster pace. Would you bet on a super-intelligent human being evil or good? If you think good is rational and evil is irrational then its obvious what the answer would be. Sure an evil person could come along and try to make malicious AI but thats also a threat with the open source AI and almost anything in the world. It also takes a tremendous amount of work to be able to manage the entire supply chain of all these products and the logistics. If AI is doing for the owners then the owner depends on the AI. I still think many things would happen before the nightmare reality. The billions on this earth wouldnt just slowly let companies swallow whole industries
Just got a Waveform coffee mug the podcast tastes so much better with it!
31:59 As someone who is familiar with Samsung's policies, I believe the initial price of the handset will be $2,200-$2,000, and then the price will drop after a couple of months until the average consumer thinks it's a bargain.
1:28:36 I believe Boom Overture will be designed to divert some of the sound energy created by a sonic boom upwards so as to reduce the noise levels of supersonic flight above land.
Yes, and Concorde did NOT fly supersonic over land, per regulations to lower sound pollution. There was no NY to CA routes for example. Which limited its usefulness, and was one of the reasons they only made about 20 of them instead of the hundreds initially projected to be sold.
Anyone wanting to know more about Concorde should read Mike Bannister’s book, called Concorde. It’s an amazing aircraft.
No wonder everybody missed the Deepseek news on Christmas Day. We had the whole H1B visa bust up at same time.😂
BTW if you remember the M.U.G.E.N engine, which involved in lots of 2D fighting games, its name in Japanese (“無限”) means the same as Moohan in Korean (“무한”). Both name derived from the Chinese word “無限”(wú xiàn), yes “infinity”.
DeekSeek's inference cost is really cheap, but it isn't clear what the actual cost of retraining or updating the model is. The localized versions of DeekSeek are basically a snap-shot of the available training data that doesn't have new information.
Another episode?? I’m loving this 🎉❤
I am so excited for Pebble coming back. I was a kid when it was first around and always wanted one. e-ink just makes so much sense for a watch, I don't need or want to be able to scroll Instagram or whatever. The Pebble just got the basic needs done right and done perfectly.
This podcast is was looking forward to
and just like that another week has gone by. Weekend starts in the UK, and as always its the Waveform to kick it off :D :D
One of the few highlights of my pitiful life. Thank you guys.
This is most neatly designed studio out of any other podcasts studios, noice
15:44 This demo of Google maps in the virtual world looks so awesome, vivid and with rich detailed simulation. I can't wait to see when Google integrates this type of simulation to make it easy for people to find their way around places. Just imagine for you to get a clear popup image of your desired location within seconds of inquiring on your phone. That will be really super awesome 😎💯💪🏾👍🏾
I just watch it for the vibe now😂❤
Marques not understanding its a preview event because its for a videogame system is hilarious to me!
53:00 Not to brag but my Garmin smartwatch has an AMOLED display, does fitness tracking, and all of the notification stuff while lasting for 5 days (27 days if I put it in power saving mode). Garmin also sells some more hardcore models with solar charging that you literally never have to plug in
I pray for imp tech news to be released before Wednesday every week so we can sit and deconstruct it together in this podcast.
Waited to do my dishes from breakfast for this.
Correction - o1 does not require a $200/mo tier, it’s always been available in the $20/mo tier as well
My notification summaries have always been pretty accurate. I do like the the italics tho
58:45 You mean like 90% of sub $50 smartbands and watches? Amazfit have a whole set of those. They even have some basic heart rate sensors, but you can disable or ignore that. It' just shows the time - and you can customize the "watchface" completely. Show everything. Or keep it minimal, like me. I have just red retro digital numbers on a black background for time. And the screen pops up when I get a notification. The battery lasts a week or so. E-inkt versions last even longer.
Also Xiaomi band pros and even a Samsung fit 3.
Happy to be here!!!!
These high speed planes are literally being promoted and funded by Saudi Arabia because they consume so much jet fuel. How can we even talk about investing high carbon technologies in the current state of the world. Honestly hope these planes never get used
Hell yes I would go on a free Boom flight!
One of my few regrets in life is putting off flying on the Concorde. It’s amazing.
I encourage you all to go on the Concorde tour at the Intrepid Museum; seeing it in person really shows how different it is compared to regular airliners.
Barely had any dishes piled up since the extra Waveform this week🔥
Been waiting for yall to respond on this!
Watching from Paraguay! Cheers David
David just did one of the most balanced overview of DS! Bravo!
If you just want a smart watch with basic functions on the cheap, why not just get a Huawei watch? 2 weeks of battery life, HR/O2sat/step tracking, notifications, music control, big screen, and pretty stylish. If you have android you can even sideload the international version of the app and activate more advanced functions like GPS, app store, watch face store, and if you have the Pro version (more expensive though) then stuff like BP monitoring and EKG as well.
Based David for the intro, I like it!
MKB Awesome Podcast.Welcome back.....Love the shirt also thank You guys for sharing the FurBaby on the Team.I also miss the Pebble Watch especially with the leather band.Thank You for sharing all updates.💯✌️
I'm still using my Pebble Time Steel and Pebble Time. Nice to know about my notifications or calls I can ignore with a glance without having to take out my phone or keep it on me at home or office. Always using it for media control and while I've cracked the glass from dropping it on cement, I've stopped even checking after scrapping against a wall because it's always buffed right out. Except for the Round they were all water resistant while Apple's first watch like Pebble's Round was only splash resistant. I'd still be using my Pebble Steel for it's style and higher contrast if I'd replace the zebra-stripe connector. I have not replace the batteries and they still last several days thanks to the Time Steel's originally 10 day battery.
From South africa really enjoyed this conversation about tech breakdown and costs amazing
No one cares where you're from, like that didn't add much to the comment...
- as a Security Guard from Poland
17:04 Interesting! Since I enjoy my Tab S10U much better than my iPad Pro, this gives me a lot of hope!
YES!!!! absolutely would want to be on the first flight.
Regarding what was being said of how Pebble should be, you ended up describing my Amazfit Bip 😂
I love how the video playing inside of android xr at 10:15 is the exact same video Ellis talks about in one of the previous episodes he watches while on facetime
Been watching your pod since October of ‘23. It’s been awhile since I first watched. Crazy how time flies ❤
how ironic that i am packing for a move, and just came across my pebble watch while listening to this LOL
I'd love a new Pebble watch. I went from Fossil to Samsung to Fitbit and now use my Casio Illuminator. If this thing buzzed and had a notification LED, it would be perfect. Also, put me on the first flight.
I'm surprised you didn't mention that one of the things companies are working on with supersonic planes is minimizing the sonic boom
Is the podcast no longer posting in audio-only format anymore? The last 2 episodes defaulted me to the video format and worries me about data consumption.
Glad Maques is back. I actually don't watch when he's not around, sorry don't ask me why!
this is one of my favorite podcasts
I JUST finished making my food to watch Episode 8 of Beast Games, then I saw this notification. JImbo can wait! 😝💯
1:25:06 he said “Mach two” and that’s a win in my eyes 😂
computerphile was my childhood
love the channel
It's the weekend already! 👀🍻
The best intro so far. I want a gif of David fake crying. ❤️
You don’t need e-ink for basic smartwatch to last a week on battery. Simple smartwatch already lasts a week or longer with OLED display
8:01 They need to stop making you look like a massive INSECT when you wear their XR headsets
Concorde was so ahead of it's time, really makes you think.
1:29:41 good call on bailing on that Blue Origin flight in the dream😂😂
I had a flight once that was delayed by a half hour before we heard over the intercom "We sincerely apologize, preflight checks revealed an issue with the aircraft so the pilot has made the decision to take a plane from the boneyard to ensure your safety. Please stay patient as we take another hour to ensure the plane is ready for flight"
So many of our planes are already edging the speed of sound -- most of my flights recently are in the 750 and 760mph range, the most I've seen is 767mph when we were in a jetstream!
None of these create sonic booms because they're being pushed by the air around them, so relative to that air it's still slower.
@44:47 The paradox you're referring to is known as Jevons Paradox. It states that as technological improvements increase the efficiency of resource use (or as the cost of a product decreases), overall consumption of that resource (or product) can increase rather than decrease.
Regarding the Pebble, Quinn from Snazzy Labs just reviewed an e-ink tablet, TRMNL, that costs $120, available now, which can display essentially anything from your phone/computer, months-long battery life. Not something I needed but ordered one this morning anyway; hope you guys can review it soon. EDIT: oh, and flying in the jetstream does NOT burn more fuel; it just isn't there all the time.
0:40 No Ellis? So just Marques and 2 not-Ellises this week. Okay, let's see.
Yes they did and Im glad they did on so many levels!! im now able to implement it into my CUDA application
People really forget Googles business model. Let’s give this google operating system access to where I’m looking whenever I’m wearing it, also where I’m not looking and with microphones. I bought and returned my Quest 2 for this same reason.
A Casio watch with e-ink display and pebble OS would be amazing!
Pebble coming back was not in my 2025 checklist
Yeah DeepSeek, BUT PEBBLE'S COMING BACK BABY! All we need now is Google to announce Google+ is coming back.
Now would be a great time for Google+
I think having pretty much all apps on play store being able to support sm fold ratio certainly helps Android xr