Tim Beckett

Coder. Writer. Interested in #AI and EdTech

[ben @contraben]

Humans are not resources.

Introducing Contra, a new professional network for the jobs of the future.

Get more creative at https://t.co/olPYDApl09

[Momentum Chaser @electricfutures]

After several delays, @DOGE has finally posted its purported savings. Why did it take so long to create a simple webpage with a 1000-row table? Who knows! Let's dig in.

Headline number: $55B saved. They list the savings per nixed contract. This should be easy to verify then.

🧵

[Michael de Adder @deAdder]

The Globe and Mail

[James Lucas @JamesLucasIT]

9. The actual size of Africa

[ToxicKhan @ToxicaBrownGirl]

Everyone rushing to X to see if chatgpt’s down #chatgpt #chatgptoutage #ai #Crypto #isitdownorisitjustme

[Santiago @svpino]

You can download this Data Science and Machine Learning book for free.

There are 4 sections for those who are starting:

• Python Primer
• Linear Algebra and Functional Analysis
• Probability and Statistics
• Multivariate Differentiation and Optimization

The rest of the chapters:

• Importing, Summarizing, and Visualizing Data
• Statistical Learning
• Monte Carlo Methods
• Unsupervised Learning
• Regression
• Regularization and Kernel Methods
• Classification
• Decision Trees and Ensemble Methods
• Deep Learning

Download the book here: https://t.co/wNK3BfAxul.

[Tim Beckett @tim_beckett]

@svpino @gree303 I'll have to try this
Find Co-pilot mostly a distraction so switched to first Cody then Codeine. For the moment still find generating code on either chat GPT or Claude then checking it myself much more efficient. Possibly a tutorial if you find a solution that suits you?

[Tim Beckett @tim_beckett]

@RpsAgainstTrump Kind of incredible to look back on this

[Dean Blundell🇨🇦 @ItsDeanBlundell]

Everything’s fine in #Toronto

[Brent Terhune in Little Rock 7.30-8.1 @BrentTerhune]

The assassination interview they don’t want you to see

[Tim Beckett @tim_beckett]

@DrewBat Also worth looking at his overall bio. He'd done some impressive things before he apparently went off the rails after his mother passed away.

[Tim Beckett @tim_beckett]

@svpino Thanks Santiago. Whats your opinion on the 365datascience platform?

[Tim Beckett @tim_beckett]

True dat

[Doug Saunders @DougSaunders]

The people who are fixated on keeping your housing expensive and your cities underpopulated are all old enough not to have mortgages

[Tim Beckett @tim_beckett]

@svpino Thanks. Ill probably start today. Curious what you have to suggest. I was about to give up on unworkable frankly. Always like your posts here so looking forward to it

[Tim Beckett @tim_beckett]

@svpino Ok cool ill have a look. 20 bucks isn't much if it can help me figure out this beast.

[Tim Beckett @tim_beckett]

Just published Sierpinski’s Triangle PT 3 https://t.co/dZikixHUjI

[Tim Beckett @tim_beckett]

Just published SierPinski’s Triangle PT 2 https://t.co/SScBprEt9y

[Santiago @svpino]

What will happen to ChatGPT if StackOverflow dies?

It's no secret that OpenAI trained their model using many data sources, including public questions and answers from StackOverflow.

But as soon as ChatGPT went live, traffic to StackOverflow disappeared. They laid off a significant portion of their team, and honestly, I don't know how they can survive this.

StackOverflow is a dinosaur, and there's a huge comet coming their way.

Many people get mad at me for saying StackOverflow is dying. They argue that AI needs StackOverflow's data, and you can't have good code-oriented models without StackOverflow.

That's bullshit.

Large Language Models need data, but StackOverflow is hardly the only place with high-quality data.

StackOverflow's advantage is that it concentrates the information in a single place, in a question-answer format. This is fundamental for people but not necessarily required for a model.

An LLM can synthesize scattered information and present it cohesively to a person. It can combine a blog, a GitHub repository, and a YouTube video and give you an answer in seconds.

What would you say if you had to bet between AI getting much better or people returning to StackOverflow?

StackOverflow solved our problem one day. But times have changed, and there's a new, much better solution out there. A solution with the potential to become exponentially better.

I don't see people coming back.

(It turns out we don't generally like places where others make us feel stupid for asking the wrong question.)

And remember, StackOverflow doesn't own that data; we do. They can do nothing to prevent models from using that data. For years, they have benefited by using our content. They can't complain because someone else is now doing the same.

We'll see what happens, but my money is on the models.

[Tim Beckett @tim_beckett]

Sierpinski’s Triangle, Fractals and Jackson Pollock PT 1 https://t.co/KgloKYlDWW