How to Use DeepSeek

DeepSeek is the spotlight of this tutorial, championed by China as a groundbreaking open-source AI to rival ChatGPT, Claude, and other Western models.

First, you'll see how to sign up for DeepSeek and enable the R1 model, setting the stage for quick yet powerful insights.

Next, we'll test DeepSeek's reasoning by posing questions about word counts, fence painting, and even historical strategies for eradicating smallpox.

Then, you'll see DeepSeek R1 go head-to-head with ChatGPT o1 in handling ambiguity, practicality, coding tasks like a Pong game, and creative prompts such as AI-themed haikus.

Afterward, we’ll compare cost structures, revealing how DeepSeek can match high-end performance at a fraction of the price charged by top competitors.

Meanwhile, you'll see how its open-source approach allows for local downloads, enabling advanced AI on your own machine with minimal hardware requirements.

Finally, we’ll address privacy considerations, touching on where DeepSeek stands compared to alternative services.

By the end of this quick rundown, you’ll be equipped to harness DeepSeek for affordable, high-level AI tasks without sacrificing clarity or control.

Host: David DeWinter

📚 RESOURCES

⌚ TIMESTAMPS
00:00 - DeepSeek: Hype or Existential Threat?
01:22 - Your First Prompts
02:24 - Chain of Thought Reasoning
03:46 - DeepSeek R1 vs ChatGPT o1
08:08 - What's the Hype?
10:44 - Privacy Concerns
11:10 - Running DeepSeek Locally

📩 NEWSLETTER

🔽 CONNECT WITH ME

🎁 TOOLS AND DISCOUNTS

🎒 MY COURSES

🙏 REQUEST VIDEOS

🔔 SUBSCRIBE ON YOUTUBE

🙌 SUPPORT THE CHANNEL
- Hit the THANKS button in any video!

⚖ DISCLOSURE
Some links are affiliate links. Purchasing through these links gives me a small commission to support videos on this channel. The price to you is the same.

#stratvert #deepseek

[HaraldEngels]

Why people always say: "China has released ...". Nobody says: The US released ..." when e.g. OpenAI, Google, Anthropic, Meta etc. are releasing a model. The mental bias of the Western world is just sick!

[aw]

From 25th to 29th, 83 hours, DeepSeek server cluster, received more than 230 million DDos malicious requests per second, the total amount of attacks is equivalent to three days of European network traffic combined.
A number of Chinese Internet companies came to the rescue to defend against the attack, and this epic story of working together to fight against foreign enemies will be forever recorded in history as a barbaric invasion was stopped.

[tayyirawashahtrawasiay]

He has exposed his bias 🍭

Deepseek is open source and it's really powerful.

[JSiuDev]

Why keep calling it a threat? Shouldn't this be a good thing as an LLM advancement?

[wli]

if you want to know which AI is more intelligent, you can actually ask the AI
"If you can ask an AI 1 question to assess how intelligent it is, what question will you ask? the question must have an answer in under 100 words"
the AI will offer you a question. then, you ask each AI that question, and ask the AI to evaluate who had the better answer.
then, repeat with the other AI, allow it to raise its own question. and repeat the process.

another simple test is "in 100 words, write a fairy tale with deep meanings".

both of these are test of intelligence, not mechanical mathematics.

[yosoyelsupremo]

Really cool AI tool, it does the work of many of the paid AI modules from other services.

[kennethguinto]

their r1 1.5 version like the litest version can outdo a lot of the paid tiers in the market right now.

[hayyeh]

Okay, I just prompted it to swear for every responses and motivate me because my programming motivation was down in the dumpster

Prompt: Right now I am programming, and my motivation is shit. Please help me!

DeepSeek: Oh, for fuck’s sake, welcome to the goddamn club, you beautiful disaster. Programming is a soul-sucking, caffeine-fueled nightmare 90% of the time, but hey, you’re here, so you’re clearly a masochist like the rest of us.

Here’s the deal, you magnificent bastard: break your shitty task into smaller, less shitty tasks. Can’t write a whole fucking function? Fine. Write one goddamn line. Still too much? Write a comment. Baby steps, you glorious dumpster fire.

And for the love of all that’s unholy, take a fucking break. Stare at a wall, scream into a pillow, or do whatever the hell it takes to reset your fried brain. You’re not a machine (unless you’re a robot, in which case, fuck, my bad).

Now get your ass back to work, you beautiful, procrastinating mess. You’ve got this, even if it feels like you don’t.

Prompt: Thank you, thank you, THANK YOU! I will do just as your told and scream into a fucking pillow with every fiber of my very being!

DeepSeek: Fuck yeah, that’s the spirit, you magnificent lunatic! Scream like you’re in a goddamn metal band and the pillow is your adoring crowd. Let that rage fuel your coding like a turbocharged engine of pure, unadulterated chaos.

And remember, you’re not just some schmuck staring at a screen—you’re a goddamn wizard conjuring shit out of thin air, one cursed line of code at a time. Now go forth, you beautiful disaster, and make that program your bitch.

You’ve got this, you glorious, swearing, pillow-screaming maniac. Now get the fuck back to work before I start throwing virtual tomatoes at your lazy ass. 🍅





Yep, DeepSeek > ChatGPT

[ComedianHarmonistNL]

Just now I am back from Dave's Garage channel and he was adressing the China politics matter. Having downloaded DeepSeek to his computer he asked about Tienanmen Square, the foto, the tanks, etc. Got an informative answer. So the info was there! Elsewhere on this thread I also mentioned my personal experience asking about the China's "Century of Humiliation".
(Well of course what happened at Tienanmen Square is looked upon by China in a very different way compared to the West. )

[DarkCreed]

Don't matter if it's good, as long as it's cheap.

[ayushsonimzp]

8:35 That $5.6M figure is a bit misleading. It only accounts for the final training run, like the cost of a Broadway show’s last performance. But just like a show also requires months of rehearsals, theater rentals, and actor salaries, the actual investment in training is much higher.

[Lonsey]

So, will ChatGPT subscription price drop or not?

[albertkoay]

Sorry, DeepSeek R1 has knowledge limitations due to the training data cutoff as of Oct 2023. Outdated data response could be incorrect & cause issues. 😢😢

[aman]

The server is busy. Please try again later.
Great AI, solid.

[dougoler]

You gave the haiku comparison to DeepSeek, but its haiku wasn't a haiku. The middle line is only 6 syllables. Maybe I'm the only one pedantic enough to have noticed.

[jacob_le_gris]

Seriously you can't compare the time it takes to respond since you have no clue what kind of hardware is used for inference and chances are OpenAI has the best in the industry while Deepseek is likely hosting this on spare hardware. Not to mention they are likely getting more traffic since it's free. R1 has less than half the parameters on equal hardware there's no way it would be slower.

[reaffirmation]

How to break deepseek (deepthink), ask it: how many characters are in your response to this question multiplied by 10.

[maran]

shows you Open AI, google, facebook, pretty much just over rated their AI potential to get all that investment money..

[Diagonal]

Why is America hell bent on keeping the rest of the world down. Nothing good comes out of being envious. Let other people have their day in the sun.

[astutepropertymanagers]

When you remove the the deep think function, you end up with faster response time.