تدريب Shadowing: Is AI Doom Real... or Just Good Marketing? - تعلم التحدث بالإنجليزية عبر الفيديو

اختر وضع التدريب

كل وضع يفتح غرفة تدريب بملء الشاشة.

180 جمل

إذا كانت الجمل قصيرة أو طويلة جدًا، انقر على Edit لتعديلها.

I am seeing a ton of doom and gloom over AI in my feed, and I wanted to make a video to balance things out

because there are certain things I want to bring to your attention.

You may know it, you may not, but at least this will give you a more balanced opinion.

I think you need to see both sides to really form a good opinion of things.

And Michael Burry just released this, and I wanted to break it down.

One, LLMs are not AI and they're not AGI.

That's true.

They're not AGI and they never will be.

Competition is coming up fast.

Check out GLM 5.3 fast if you're not familiar.

It's very competitive.

It's very close to the frontier models that we're currently seeing from Anthropic and OpenAI.

That's impressive for an open weight, freely given model.

IPOs and hype and puffery.

Ah, it's so awesome.

It's going to destroy the world right before an initial public offering,

right before all the stock becomes available for you to buy.

Ah, maybe to hype up valuations a little bit, maybe.

And cover for uncontrollable slowing growth.

Is the market saturating?

Are we getting to the spot where they can't grow any bigger

and you're going to start to see that curve flatten out.

And then they're like, oh crap, we're not at a trillion dollar valuation.

We're closer to like 60 billion.

That's a big gap to cross.

But before we break down into the next bit, I need to pay some bills.

So check out our sponsor.

Have you ever had GitHub runners take forever?

Well, when they're working, that is.

I will say I switched over to Blacksmith

and I went from about a five and a half minute run to blacksmiths one, which was about two minutes and 50 seconds.

And then all the way down to about a minute and a half, more than probably two, even up to three X as fast as the built-in GitHub runner.

And migrating was really easy.

You can just go to the migration wizard, select your repository, hit next.

It scans all your runners says, hey, we found two on GitHub's Windows latest.

We'll migrate those over to blacksmith just next next and then generate the pull request for me.

It'll go ahead and switch those over and use blacksmiths.

I'm going to go ahead and squash and merge this.

So I will get back on there because I don't want my runs taking five minutes.

I much prefer the one minute one with blacksmiths cache and faster runners.

And if you'd like to check it out today, go to link.christitis.com

forward slash blacksmith now i'll show this tweet by sam altman

where he's basically calling for regulation of american companies at the

frontier model uh essentially this is just sam altman dario from anthropic

and then also elon musk of xai and grok um google If you ever use Gemini,

you know they're not really part of the conversation.

But anyways, breaking down this post, one thing that sticks out to me, it's weird when someone calls for them to regulate themselves.

It's very strange, especially at this level.

Because anybody that's ever done a government contract knows how much bureaucracy and red tape it is.

It slows you down drastically and it increased costs across the board.

So why would someone ever want regulation to hit themselves?

Is it because maybe they're afraid it's going to get out

and Skynet's going to launch the missiles tomorrow

or is there another point of view to this i think it's somewhere in between

but i want to first lay out the other point of view

because i look at this and i just every gut instinct in me says it's

and i i look at it and it's like hey we're not stopping we're pacing

and the reiteration of American competitive pressure and those types of things that American companies

leads me to believe they're going to try and lock out all open weight and Chinese models.

I don't know how they would do that, but it feels like they're coming up too close and they might eat their lunch.

And they're way cheaper than paying the $200 to Anthropic for their subscription when you could just use the cheaper stuff from China.

And it's almost as good.

So that's where we are today now imagine a year or two with a ton of efficiency built in,

maybe you get to the point where even local models become good enough.

If local models follow their current trajectory, the local models a year or two from now might be as good as the frontier models are today.

You never know.

I mean, just saying, I don't think it's going to get to that point, but it might, or I think it'll get to the point where it's good enough for 99% of the users,

so they don't need to pay OpenAI or Anthropic.

I think that's a real possibility.

So that's where we stand from them.

And then you start to look at like some of the valuations.

I love the clip from the big short here.

Anthropic valued almost at a trillion dollars above Walmart,

but their revenues are less than a tenth of Walmart where they're saying they're making 65 billion. That doesn't make sense.

Totally normal, right?

And then another person looked at the Anthropic finances And this struck me as strange,

but you might call for your own regulation if you also own or have vested interest in the regulator.

So Meter, the model evaluation threat research company, well, they have a massive stake in Anthropic.

And currently, they're a third party evaluator to assess the risks of Anthropics models.

So if they do get regulated, does that mean Anthropik's double dipping?

To an extent, because that regulator is regulating an entity where they want to succeed because that means they get more money.

But if they come out and say, oh my gosh, this company's terrible, we need to slash them, we need to lock them down,

and then they lose a lot of the revenue, well, that means they would also lose almost all of their money because they have,

what, 500 million stock in Anthropic last year, currently 7.7 billion today, 16 months later.

That's insane.

So if they come out really aggressive against Anthropic, they would essentially just be burning and throwing away $7 billion worth of stock in anthropic.

So that kind of makes more sense now, doesn't it?

I would want someone to regulate me too, if that means I can make even more money.

One, I get money from you for signing up for a subscription.

But then I get money from you again, because you hear taxpayer money is coming from the government.

I'm getting a double dip from that section.

And now all of a sudden, oh, the regulation starting to make a little more sense.

Now, this is pretty conspiratorial.

This is not factual.

I'll leave a link to this so you can go through the entire thing.

But this guy went to pretty extensive research to say, hey, this could be happening behind the scenes.

And it makes me pause and go, that makes more sense than Skynet launching the missiles.

When I look at LLMs and their capabilities, this makes more sense than the other opposing point of view.

I'm not saying it's right or wrong.

100

I'm just saying it does pass that sniff test to me.

101

And then also, like for the younger audience, I know a lot of people are hopeful for government regulation that they'll do the right thing.

102

But I'm in my mid 40s and I just kind of know better.

103

I remember the 90s and 2000s.

104

I remember in the 90s this, the section 230 and how social media companies abused this, essentially being saying, hey, if someone uploads something to our platform,

105

we're not responsible for it.

106

We can't be held for all the stuff that was uploaded.

107

They hide behind this get out of jail free card.

108

And that's part of regulators protecting big tech.

109

And there's other instances of self-dealing, just a gajillion of them out there.

110

I'm not going to talk on Enron and Halliburton, Dick Cheney and all that from the early 2000s.

111

That's a fun one if you ever want to go down that rabbit hole.

112

But to stay on topic here, what are the real risks?

113

Biological warfare that could develop weapons?

114

I mean, one, why are the models being trained with this information?

115

That's one thing I'd be like, well, how did it get in their model?

116

That's kind of odd like models should be trained on what is actually being used.

117

So why is that even public knowledge or publicly accessible?

118

Cyber attacks.

119

I think this is possible with enough training.

120

Like you look at it finding like a 20 year old Linux bug, I think after like $100,000 worth of inference,

121

there's still a huge amount of compute tossed towards that bug finding, but it could still be you string enough of those together, it could be an attack surface.

122

So I get that as well.

123

And I look at this last one too, where it's like the loss of control.

124

When you think about the Hugging Face hack where OpenAI's model hacked into Hugging Face

125

and cheated on the test so it could score higher on the benchmarks.

126

And I'm thinking if I'm benchmarking an AI, why does it have access to the internet?

127

Why isn't it sandboxed properly?

128

And I know the hugging face hack it was apparently sandbox but it got out through some kind of exploit

129

and i'm thinking to myself as a sys admin like you could put this on a dedicated network

130

that has no internet access like even a vlan would be very difficult to bypass

131

if it's hard-coded in like a cisco router or something like

132

that so i'm kind of sitting here going to to a normal person i guess

133

that makes sense but to me i look at it

134

and i go

135

that just seems like incompetence on the sys admins setup it

136

seems like the lab environment was not set up properly for it to to be able to exploit its way out.

137

Those should be like hard-coded limits.

138

And heck, it's not even just like a physical network cable that could be, you know, disconnected essentially.

139

But I mean, you could go that route, but I don't even think you need to do that.

140

A lot of this is virtualized and you'd have a virtual subnet

141

and that could be locked off in its own environment from the VM level.

142

And there's no getting around that.

143

I just don't think there's enough exploits in the world to get around that.

144

And I just, I look at things like this and I just suspect.

145

I look at what the incentives of these companies are.

146

And anytime I see a company say, I want to regulate myself, I immediately go, what is going on here?

147

This doesn't pass the smell test at all.

148

And I want to put this out there.

149

And I know a lot of people are really like AI doom and gloom.

150

And I think a lot of the Frontier models are great.

151

I use them all the time.

152

I think they can be good for programming.

153

But a lot of this seems like,

154

hey, there's not enough of me out there buying a $200 subscription to rationalize the $1 trillion valuation.

155

99% of the population, honestly, is going to be able to use a local LLM or a cheaper model elsewhere and get all the benefits they need.

156

They don't need these big models.

157

And that's the scary part.

158

Because when this thing crashes, I know a lot of people are hoping for an AI bubble crash.

159

And I honestly am not.

160

Because I look at my portfolio, I look at my 401k, my retirement is someone in his 40s.

161

Everything's infected with it.

162

Every broad market index fund is just heavily invested in it.

163

Heck, I even looked at putting everything in like a materials index fund

164

and come to find out as soon as I dive deep into those Vanguard index funds from the material side of things, some of them are in like gas companies.

165

I was like, oh, okay, German gas company.

166

That's pretty, you know, diversified.

167

And then come to find out they supply like 80% of the gas to SpaceX, which XAI is part of.

168

So if AI bubble pops, XAI, SpaceX has to absorb that.

169

Maybe they cut out a bunch of launches and then all of a sudden gas demand goes way down.

170

I mean, there's just no getting around this thing.

171

And when this thing does, if it does pop, I don't know, maybe the dollar just deflates into nothing.

172

I don't know.

173

I could just rant on and on and on.

174

I just wanted to make this video because I saw this point of view and I thought this is interesting.

175

I haven't seen anybody talk about this.

176

And I'm thinking, man, this would make a good video.

177

So let me know, was this informative?

178

It's a little bit different than my usual, but let me know down in the comments and I'll see you in the next one.

179

Next one, I've got something special I've been working on for a while now and just a sneak peek.

180

Anyways, see y'all later.

تنزيل التطبيق

كل ما تحتاجه للتحدث بطلاقة

تقييم النطق بالذكاء الاصطناعيتقييم كل جملة

تدريب IPAأتقن كل صوت

المفرداتابنِ رصيدك من الكلمات

لعبة المفرداتتعلّم أثناء اللعب

امسح رمز QR

Download on theApp Store Get it onGoogle Play

نفس القناة

I Rebuilt My Entire Website in Astro… Was It Worth It?

Daily ConversationWorking Adults

I Spent 1.5 Billion AI Tokens on Absolute Slop

Business EnglishUniversity StudentsWorking Adults

FOR YOU

فيديو موصى به

63 جمل

المفردات وملاحظات التحدث لهذا الدرس

يحتوي هذا الفيديو على 180 جملة و2272 كلمة للتدرب على الترديد. مدة الكلام 13:48. يتكلم المتحدث بسرعة طبيعية، نحو 165 كلمة في الدقيقة، قريبة من المحادثة اليومية. 81% فقط من الكلمات ضمن أكثر 3,000 كلمة شيوعًا في الإنجليزية، فالمفردات صعبة نسبيًا.

أهم المفردات في هذا الفيديو

15 كلمة من الفيديو تستحق التعلّم، مع النطق والمعنى:

الكلمة	النطق	المعنى
regulation اسم	/ˌɹɛɡjʊˈleɪʃən/	تَنْظِيم
regulate فعل	/ˈɹɛɡ.jə.leɪt/	ضَبَطَ
runner اسم	/ˈɹʌnɚ/	عَدَّاء, عَدَّاءَة
frontier اسم	/ˈfɹʌntɪə/	حَدّ, حُدُود
hug فعل	/hʌɡ/	عَانَقَ, اِحْتَضَنَ
subscription اسم	/səbˈskɹɪpʃən/	اِشْتِرَاك
exploit اسم	/ˈɛksplɔɪt/	مَأْثُرَة
regulator اسم	/ˈɹɛɡ.ju.leɪ.tɚ/	منظِّم
sudden صفة	/ˈsʌd.ən/	مُبَاغِت, مُفَاجِئ
bubble اسم	/ˈbʌbəl/	فُقَّاعَة
missile اسم	/ˈmɪs.aɪl/	قَذِيفَة, صَارُوخ
upload فعل	/ˈʌpˌloʊd/	حَمّل
doom اسم	/duːm/	نَحْس, شُؤْم
topic اسم	/ˈtɑpɪk/	مَوْضُوع, مَوَاضِيع
extent اسم	/ɪkˈstɛnt/	اِتِّسَاع, مَدَّى

الأفعال المركبة التي ستسمعها

الكلمة	النطق	المعنى
come out فعل	/ˌkʌm ˈaʊt/	خَرَجَ, غَادَرَ
come up فعل		طَلَعَ
go down فعل		نَزَلَ
throw away فعل		طَرَحَ

القواعد في هذا الفيديو

أكثر التراكيب التي يستخدمها المتحدث، مع العبارات نفسها من الفيديو:

التركيب	في الفيديو
المبني للمجهول be + اسم المفعول — التركيز على ما يحدث لا على من يفعله	can't be held · being trained · should be trained
المضارع التام have/has + اسم المفعول — حدث ماضٍ ما زال له أثر الآن	Have you ever had · have vested · haven't seen

نطق يحتاج إلى انتباه

يستخدم المتحدث 59 من الصيغ المختصرة، مثل they're, I'm, don't. انطقها بصيغتها القصيرة كما تسمعها.

أصوات “th”: tenth /tɛnθ/, anthropic /ænˈθɹɒp.ɪk/, blacksmith /ˈblæk.smɪθ/
أصوات “sh” و“zh”: regulation /ˌɹɛɡjʊˈleɪʃən/, essentially /ɪˈsɛnʃəli/, valuation /ˌvæ.ljuːˈeɪ.ʃən/, subscription /səbˈskɹɪpʃən/, benchmark /ˈbɛn(t)ʃmɑːk/
الكلمات الطويلة — انتبه لموضع النبر: regulation /ˌɹɛɡjʊˈleɪʃən/, essentially /ɪˈsɛnʃəli/, valuation /ˌvæ.ljuːˈeɪ.ʃən/, regulator /ˈɹɛɡ.ju.leɪ.tɚ/, efficiency /ɪˈfɪʃ.ən.si/

كيف تتدرب بهذا الفيديو

استمع إلى الفيديو كاملًا مرة واحدة دون أن تتكلم، ودوّن الكلمات التي لا تعرفها.
ابدأ بسرعة 0.75×، وردّد جملة جملة، ثم عُد إلى السرعة العادية عندما يصبح الأمر سهلًا.
سجّل صوتك وقارنه بالأصل، مع الانتباه إلى كلمات مثل regulation, regulate, runner.

ما هي تقنية التظليل الصوتي؟

التظليل الصوتي (Shadowing) تقنية تعلم لغة مدعومة علمياً، طُورت أصلاً لتدريب المترجمين الفوريين المحترفين. الطريقة بسيطة لكنها قوية: تستمع لصوت إنجليزي أصلي وتكرره فوراً بصوت عالٍ — كظل يتبع المتحدث بتأخير 1-2 ثانية. تُظهر الأبحاث تحسناً كبيراً في دقة النطق والتنغيم والإيقاع وربط الأصوات والاستماع والطلاقة.

تقنية الشادوينغ: اقرأ الدليل الكامل خطوة بخطوة →

تدريب Shadowing: Is AI Doom Real... or Just Good Marketing? - تعلم التحدث بالإنجليزية عبر الفيديو

اختر وضع التدريب

تنزيل التطبيق

I Rebuilt My Entire Website in Astro… Was It Worth It?

I Spent 1.5 Billion AI Tokens on Absolute Slop

5 open source tools that replaced my $320/mo AI stack...

Introducing Become an Epic Product Engineer

Something Feels Off About 2026...

AI Ads Are Out of Control

He’s always wrong

Has AI killed Tailwind?

Modrinth Just BANNED AI Mods

I tried gaming on Linux...

CI/CD Explained in 5 Minutes

7 Programming myths that waste your time

Claude Just Revealed AI's Biggest Problem

Introducing GPT-6 Astra: the most intelligent and aligned model in the world.