Francis Yates

Senior Deep Learning Research Scientist at @3M Dublin. Engineer with a passion for robotics. I work on 3D mesh Transformers. All views are mine

[Yann LeCun @ylecun]

Most of these signatories have a distorted view of what is coming next with AI.
The distortion is due to their inexperience, naïveté on how difficult the next steps in AI will be, wild overestimates of their employer's lead and their ability to make fast progress, and financial incentives to hinder open source AI platforms (since almost all of them work for providers of proprietary AI systems).

As you and I know, generations after generations of AI researchers have made the same mistake: thinking that human-level AI is "just around the corner" because of some new paradigm they are working on.
They have consistently underestimated how difficult it is because they have an incentive to dismiss the limitations of their favorite paradigm as mere "engineering problems," and they can't foresee the obstacles they'll bump into before bumping onto them.

We are making progress towards human-level AI, but we're still far from it.
Some of us may have vague roadmaps for how to get there. But no one has a credible blueprint, let alone a demonstrable prototype.

Before such a prototype exists (perhaps with the learning abilities of a house cat), regulating AI R&D because of a fear of existential risk is highly premature.

There is nothing wrong with sensible legislation to regulate the *deployment* of AI application.
But legislation that hinders open research and open source AI platforms, and that makes open source AI developers liable for what people do with their code is extremely regressive.

[Francis Yates @considerahead]

@LukeDaly1979 @Feljin_J Sure, but they're still used by loads of people.

[Francis Yates @considerahead]

@LukeDaly1979 @Feljin_J The section from 5 lamps to town also makes the commute safer for cyclists, scooters, etc. So it’s not exclusive to Clontarf commuters.

[Francis Yates @considerahead]

@LukeDaly1979 @Feljin_J When I was a student, I cycled that route everyday, saving 3 euro ish per day on the bus. It was like 5km, so too far to walk in the morning. Cycling was faster than the bus too. The extra 20 quid per week made a difference. The car traffic was exactly the same back then, too.

[Francis Yates @considerahead]

@gallabytes @ESYudkowsky @ylecun In response to this thread, the critical first try is something unbelievably rare in engineering. Almost all progress in the 20th/21st has been iterative and deliberate. Seldom do we we have light switch moments as @ESYudkowsky suggests. Even in AI field, we've had almost none

[Brian Mac Mághnais @TheBrianMcManus]

I try my best to not talk about Irish politics, but getting sent things like this by my Garda friends just angers me too much. This happened today.

The police (Gardai) in Ireland have been attacked and bullied by scumbags for years now. They know they can get away with it.

Stripped of any power to do anything about it. The Gardai are trained to retreat and not engage when violence breaks out.

This is what happens when you remove your police forces ability to police.

The Gardai voted with an enormous majority vote of no confidence against the head of the Gardai. Our minister for justice ignored them. Overworked, underpayed, and stripped of any protection in policing.

I know policing in America is a sore topic for good reason. Ye are on the other side of the spectrum. Too much power. This is what happens when you swing to the other side. Chaos.

I’m ashamed of how Ireland is being governed. A failed socialist state. It’s why I left, and it’s why I won’t be going back any time soon.

Helen McEntee, you need to resign.

[Francis Yates @considerahead]

@billpeeb Great talk at ICCV! Such a interesting architecture. 🤗

[Francis Yates @considerahead]

@dara_moran @crazyhouseprice So people that rent have to rent from private landlords? Private landlords in my experience are worse than renting from a company.

[Francis Yates @considerahead]

@cormacDublin @DublinCommuters Have you been to London recently? The bus only lanes make is just as fast as the underground for a lot of routes.

[Francis Yates @considerahead]

@AlphaSignalAI Or the progress is due to an increasing number of people moving into the field?

[Francis Yates @considerahead]

@marktenenholtz For a thought experiment, take the weights run a few epochs fine tuning on a slight different dataset. are the weights still non-commercial? You might get 99% of the performance but are the weights considered the same?

[Francis Yates @considerahead]

@wightmanr When you’re training multiple models or doing hyper parameter search, what libraries do you use to track and manage these training runs?

[Francis Yates @considerahead]

@tunguz I use both professionally and I’m still waiting to see what PyTorch does significantly ... although I do love the hugging face integration of PyTorch.

[Francis Yates @considerahead]

@NannaInie Depends if you talking about the model GPT3 or the trained on language GPT3. plenty of uses for the architecture.

[Louis Kirsch @LouisKirschAI]

Emergent in-context learning with Transformers is exciting! But what is necessary to make neural nets implement general-purpose in-context learning? 2^14 tasks, a large model + memory, and initial memorization to aid generalization.

Full paper https://t.co/yyp9467WgF

🧵👇(1/9)

[Francis Yates @considerahead]

Just watched a great interview with Monica Daley @birdBiomechy , super interesting.
https://t.co/0we1VZxE68

[Francis Yates @considerahead]

Next week I'm going to be joining @3M as a Senior Research Scientist for AI in dental applications. Can't wait, sad to be leaving @VisualAIPeople after 4 great years.

[Francis Yates @considerahead]

@crazyhouseprice Taxed such that sale is forced. The price would be right at some point. If they can define RPZ they could do for derelict building too.

[Francis Yates @considerahead]

@sh_reya A couple years ago Tensorflow changed a kwarg in a dropout layer from 'drop_probablity' to 'keep_probability' Suddenly overnight all my models trained without error but didn't perform. I was baffled for days. Positional args 0/10 do not recommend.

[Francis Yates @considerahead]

@PhilipMcNally AI comes in clutch again