Robert E. Guinness

because I'm not graced with the privilege of long-form tweets.

4/ whatever name people have come up with for this blip in cultural history. I probably still have a myspace account, but I doubt even @myspacetom himself could recover my login creds, which I've long forgotten. P.S. I won't miss having to break up posts into multiple tweets

My experience with X, aka Twitter. 🧵 1/ I first heard about Twitter at a conference back when it was still an SMS-based service. I thought it was weird. I joined it in January 2008 when it was a website and getting more mainstream. For most of the past 15 years, I tweeted

@aussielunix @davefarley77 Isn't that a technical skill?

@davefarley77 scientific thinking - lean.org/the-lean-post/…

What NON-TECHNICAL skill do you see as the most important in making a career in software engineering? 👇

@davefarley77 Compassion

@greg16676935420 She must be from a rich family since her parents could afford 3 vowels.

I’m literally shaking right now

@WarintheFuture Hi, I enjoy your tweets, but copying content from earlier threads & swapping the combatants around is not a good look, especially when you forget to do some of the swapping. Maybe just link to earlier threads on similar topics instead of self-plagiarism? bit.ly/3ycSwjH

14/ A third consideration for the Ukrainians will be disrupting the Russians ability to interfere with the withdrawal. We should expect to see increased air defence, jamming & artillery used by the withdrawing Russians, as well as greater air support.

The Battle of #Bakhmut has now raged since August 2022. The massive human and material resources expended by Russians on this objective may soon necessitate a Ukrainian withdrawal. 1/25 🧵

Educating a team slows you down for a week or two. Not educating the team slows you down forever. Time spent in learning is never wasted.

{\__/} ( • . •) / >📦 npm i {\__/} ( • - •) />📦 oh noooo 📦📦 📦📦📦 📦📦📦📦 📦📦📦📦📦 📦📦📦📦📦📦 📦📦📦📦📦📦📦 📦📦📦📦📦📦📦📦 📦📦📦📦📦📦📦📦📦 📦📦📦📦📦📦📦📦📦📦 📦📦📦📦📦📦📦📦📦📦📦 📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦🐛📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦📦

@marktenenholtz Could we use some AI? (true story)

Offend a data scientist in 5 words or less:

@tomgoldsteincs 2/ The best way to detect plagiarism is to assess students' abilities in environments where they can't cheat, & then when they do longer assignments where they could cheat, you must scrutinize "outliers" compared to the earlier assessments. Watermarking could be 1 signal to use.

#OpenAI is planning to stop #ChatGPT users from making social media bots and cheating on homework by "watermarking" outputs. How well could this really work? Here's just 23 words from a 1.3B parameter watermarked LLM. We detected it with 99.999999999994% confidence. Here's how 🧵

@tomgoldsteincs 1/ From my short experience as a lecturer, students who are prone to plagiarizing will probably find ways around this. It might make their job of plagiarizing more difficult, but they'll use real-world "reinforcement learning" to avoid detection.

Twitter ad placement AI: I have enough t-shirts. I've never bought a t-shirt after seeing an ad, and I probably never will. Thank you.

@trishume What percentage of the people who criticized this idea actually clicked the link and read past, "I want to be clear this is meant as educational fun, and not as a good idea, at least going all the way to one machine."? Thanks @trishume for the interesting research and fun read!

I argue that a simplified production Twitter could run on one (big) machine using modern hardware like NVMe, 100Gbps NICs, A100s and a terabyte of RAM. Even typical image serving and basic algorithmic timelines might fit: thume.ca/2023/01/02/one…

@rdicosmo @ioanamanol True. Maybe I misunderstood your original question. Is your question about creating tech that can discern fact from fiction?

@RobGuinness Humans can also generate bogus content very well alone (trolls, scams, cheats have been there for a while), but with machines at their fingertips the scale changes. We need countermeasures. E.g. see @ioanamanol's work icfca2021.sciencesconf.org/resource/page/…

[thread begins] A few consideration on #ChatGPT are in order. First of all: *yes*, it is an impressive technical achievement. But I am far from convinced that this will be "at the service of humankind", unless we find a way to build proper countermeasures. Let me explain. 1/6

@rdicosmo Great Qs, but I think we'll need to first agree on a definition of "bogus content"? Does it mean any content generated by a machine/AI? Or does it mean any content portraying itself as generated by a human but actually generated by a machine? Or something else?

But where is the technology that *protect* us from bogus content? This is what I would like to see! Our AI/ML friends, any takers? 6/6

I would like a refund, please. fortune.com/2022/12/01/onl…

Scary.