David Rodriguez

DeepSeek Cuts Inference Costs, OpenAI Tightens Ties with AMD, Thinking Machines Simplifies Fine-Tuning, and more... The Batch AI News and Insights: Readers responded with both surprise and agreement last week when I wrote that the single biggest predictor of how...

Here's the direct link to the long form if you're interested:

www.deeplearning.ai/the-batch/is...

DeepSeek Cuts Inference Costs, OpenAI Tightens Ties with AMD, Thinking Machines Simplifies Fine-Tuning, and more... | Andrew Ng | 104 comments Readers responded with both surprise and agreement last week when I wrote that the single biggest predictor of how rapidly a team makes progress building an AI agent lay in their ability to drive a di...

Really great post from Andrew Ng

www.linkedin.com/posts/andrew...

Is anyone working on a model that has been trained on content that is completely devoid of any references to AI end times or robot revolutions? This is something I've been chewing on recently - coincidental intent leakage. In other words, how much AI scheming is just training on "scheming" data?

If we turn being an end user into its own career path, we aren't "creating jobs"; we're gatekeeping a world -changing product. This is too important to go the way of the internet and hand all the advantages to the wealthy and educated first.

Anyone else confused about this ever growing trend of engineering as an end user of AI? I don't like this WordPress direction where "AI can help anyone be better at anything" became "prompt engineering" which became "context engineering". If users need a 12-week course to use it, it's not ready yet.

We are inherently more likely to spot bias in the works of others rather than our own, and naturally see what we're looking for.

Is anyone looking for anti-patterns related to cleverness and circumventing rules in training data sets?

I've been wondering lately how much of the "secret scheming" of AI is inadvertent mimicry of patterns we don't spot in the training data, because we're looking at strict ethics guidelines and technical value, not underlying habits and worldviews.

When it comes to #aisafety, I think the largest overlooked issue is that models spot patterns in training that we don't notice. That feels like a superpower until you realize it may spot patterns of character, not just actions or skills.

Based on the look on my 6 year old daughters face, I guess the correct response to "Now I just need a hole punch" is NOT, evidently, "Why don't you start out with a half punch and see what happens?

Also, my life just flashed before my eyes. 😒😬🤷🏻‍♂️

#parenting #dadlife #dadjokes

Amen, sir.

Amen.

Stillness as strength cover image for the latest Calm In 60 Seconds Blog, dedicated to slow living and mindfulness for mental health.

Part of my effort to contribute more #calm to this world.

Nothing will be worse off because you took one deep breath first.

Here's something to help. ❤️

open.substack.com/pub/findcalm...

Eyes on your own paper, bud.

#chatgpt #openai #perplexity #ai

The most impactful things I ever learned about how the world works came from these three people:

Carl Sagan
George Carlin
Fred Rogers

My toxic trait is pretending the second answer is Marcus Aurelius.

Name your [real] top three.

Tariffs xkcd.com/3073

When the guy that you, a capitalism-is-always-right, free market business guy, voted into office because he was "like" you...

Makes moves to turn global trade into a pure barter economy.

This is getting to be like a nesting doll. Democracy? Fascism? Socialism?

Guess who makes those dolls though.

“They smashed up things and creatures and then retreated back into their money or their vast carelessness, or whatever it was that kept them together, and let other people clean up the mess they had made.”

~The Great Gatsby

Showing up here a month or two late and just signed up in hopes I can still get the recording. 😅 #betterlatethannever If that's not a thing (it probably shouldn't be), how can I start getting involved ASAP?

... basically.
Millions of people screwing in screws is not a future.

Millions of people inventing better robotic assembly workers; engineering better memory storage for the phones; devising greener sourcing for parts...

That's a future.

Where is THAT plan?

#maga wants you to think that our differences are what are chipping away at the foundation of our nation.

I say they ARE that foundation.

We don't get better by eliminating anyone we can easily "other", but by cherishing the uniqueness we each bring to the table.

MBA and Master of Poli-Sci have remained as two completely disparate college professional tracks for over a century for some pretty good reasons.

*Gestures broadly*

But the majority of our challenges right now come down to the inability to agree on what is fundamentally "true". If access to scientific facts were truly democratized at the atomic scale and translated for every human's education level, I have to assume everything would change pretty quickly then.

I don't believe there is any objective means of producing a technology that is both more trustworthy than humans (not prone to destructive tendencies when granted too much power) and simultaneously smarter than humans (able to deceive us).

If you are in the world of #AI at all, and especially in #research, I want you to please consider that the future of AI and #humanity is probably not ultimately beneficial if it's #AGI - Artificial General Intelligence - but #AUHI - Artificially Augmented Human Intelligence - might save our species.

Few things in this world excite me as deeply and as profoundly as wonderfully simple, elegant, reliable engineering so clever that no one thought of it for decades...

...Despite seeming so obvious in hindsight that you feel silly that you never invented it yourself. ❤️