Alan Napier (@Alanapierreid) profile picture

Alan Napier

@Alanapierreid

Irish data scientist and linguist based in Berlin. Working on ML tech for legal sector. Opinions are my own. Pazza Inter amala 🖤💙.

 Berlin, Germany
Joined May 2010
  • Tweets683
  • Following623
  • Followers80
6Media
Photo by Alan Napier (@Alanapierreid)Photo by Alan Napier (@Alanapierreid)Photo by Alan Napier (@Alanapierreid)Photo by Alan Napier (@Alanapierreid)Photo by Alan Napier (@Alanapierreid)Photo by Alan Napier (@Alanapierreid)
Inter’s problem is that they have been completely Lautaro, Dimarco and Zielinski-reliant for 4 months. With Lautaro’s injury and drop-off of the latter two, there are no top performers left on the pitch. Sommer, Barella, Thuram, Bastoni and Mhky have simply been awful all season
2
41
I've read a ton of research papers this year. And to conclude this eventful year in AI, I've compiled a selection of 10 noteworthy papers from 2023 that I am discussing in my new article: magazine.sebastianraschka.co… To get a sneak peak, I'm covering: - insights into LLM training runs - new openly available LLMs - efficient finetuning methods - improving "small" LLMs - pretraining on domain-specific data - new techniques to align LLMs with human preferences - enhancing LLMs with high-quality data and instruction sets - ConvNet vs vision transformer comparisons - recent developments in image segmentation - video synthesis with latent diffusion models I'm wishing you all a great start to 2024!
Ten Noteworthy AI Research Papers of 2023

Ten Noteworthy AI Research Papers of 2023

This year has felt distinctly different. I've been working in, on, and with machine learning and AI for over a decade, yet I can't recall a time when these fields were as popular and rapidly evolving...

magazine.sebastianraschka.com
32
299
17
1,273
149,077
Alan Napier retweeted
elvis (@omarsar0) profile picture
elvis
@omarsar027 Dec 2023
Nice work surveying 300+ papers and summarizing research developments to look at in the space of Generative AI. It covers computational challenges, scalability, real-world implications, and the potential for Gen AI to drive progress in fields like healthcare, finance, and education. arxiv.org/abs/2312.10868
Image from elvis (@omarsar0)
16
231
16
973
341,017
Trust, but Verify @ShreyaR presents a case for overcoming inconsistencies and brittleness by using @guardrails_ai when working with LLMs. Learn the techniques in her full talk here 👇 ai.engineer/summit/schedule/…
1
5
15
2,784
Absoluter Wahnsinn. Unsere Regionalbahn von BER nach Ostkreuz ist ZWEI STUNDEN auf der Strecke stehen geblieben. ZWEI STUNDEN im Wald stehen bis wir dann zum Flughafen zurückgeschleppt wurden. #DB #deutschebahn #ber #berlin
1
I have never experienced anything like this having lived in 4 countries and travelled all over the world
Stable Diffusion is pretty horrible
This quoted post is unavailable.
Try viewing anyway · View on X
2
GPT-3 has over 170 billion parameters, Here are my top 3 favorite 1/
28
200
17
2,469
Alan Napier retweeted
E L H O T Z O (@elhotzo) profile picture
E L H O T Z O@elhotzo18 Aug 2022
Leute studieren 10 Semester Germanistik nur um „kafkaesk“ zu sagen, wenn sie ein Formular ausfüllen müssen
30
70
3
3,616
Alan Napier retweeted
Dr. Tech (@doctechmd) profile picture
Dr. Tech
@doctechmd14 Aug 2022
When the project is not ready, but the client wants a demo
677
15,163
1,579
105,609
Data scientists often forget: Your job isn't always to outperform humans. Often, it's just to automate a boring task to free a human up to do something more productive. Challenge yourself to think more about utility than pure accuracy.
12
114
5
833
Alan Napier retweeted
Bojan Tunguz (@tunguz) profile picture
Bojan Tunguz
@tunguz10 Aug 2022
Be careful what you pip install. 10 malicious Python packages exposed in latest repository attack – Ars Technica arstechnica.com/information-…

10 malicious Python packages exposed in latest repository attack

Supply-chain attacks are moving GitHub toward digitally signed packages.

arstechnica.com
20
3
77
Docker uses almost 9GB of memory on the M2 MacBook Air with *zero* containers running 😬
Image from Ian 🫠 @iansu@cosocial.ca (@iansu)
96
91
34
1,167
I’m sorry but this argument from Elon’s lawyers is borderline statistically illiterate. It doesn’t matter at all that 100 accounts is a small percent of the total user base.
Image from Kareem Carr, Ph.D. (@kareem_carr)
77
270
115
2,403
Alan Napier retweeted
Mark Riedl (@mark_riedl) profile picture
Mark Riedl@mark_riedl16 Jul 2022
Musk has apparently never heard of statistics. Concerning for someone who professes to be a genius engineer
David Sacks (@DavidSacks) profile pictureDavid Sacks
@DavidSacks16 Jul 2022
Twitter is toast.
Image from David Sacks (@DavidSacks)
14
19
1
206
Alan Napier retweeted
Santiago (@svpino) profile picture
Santiago
@svpino8 Jul 2022
Many people new to machine learning have no idea that labeling data is a problem they need to think about. To be clear: "labeled datasets" aren't a thing in the real world. Here is an excellent approach to getting past this problem: 1 of 11
36
185
14
1,114
Replying to @Hoont
As a Canadian - with a culture closest to that of America - I’d like to make an apples-to-apples comparison. Let’s see if you can determine the primary determining factor in gun-related deaths.
Image from WhatTheEver (@JustSomeRandoEh)
522
3,875
207
30,128
From a genuine expert
Richard Socher (@RichardSocher) profile pictureRichard Socher
@RichardSocher15 Jun 2022
It's so hard to balance excitement for the real progress that's currently happening without extrapolating too far. It's amazing LLMs can fake a conversation so well with the objective of predicting the next word. They're def. not sentient or conscious.
1
Ranked in reverse order of difficulty: 4) SQL 3) TensorFlow 2) PyTorch 1) Setting up a meeting in MS Teams
1
You see a lot of claims about "zero-shot" or "one-shot" learning these days, used to suggest that foundation models show a degree of generality. This is highly misleading. Here's why.
13
103
13
644
Load more