đź—“ I'm not buying a 2024 calendar until I see the trailer
Late to the Party 🎉 is about insights into real-world AI without the hype.
Hello internet,
Happy New Year! And what a year 2023 was. I am really looking forward to 2024 and what it has in store!
The Latest Fashion
Three recaps of 2023 from big tech:
- Google about their “groundbreaking” advances in AI
- Microsoft reflects on 2023 in their AI advances and discoveries
- And here’s Meta’s recap about VR and AI
Got this from a friend? Subscribe here!
My Current Obsession
I haven’t done New Year’s resolutions in a while. But I usually do set long-term goals I want to achieve in the year and a theme for the year.
My last year was “Focus”, but I feel like focusing actually wasn’t really my style. Taking opportunities that pop up along the road is what led to writing this newsletter to over 1,000 people, among many other things.Â
So in 2024, my theme is going to be “Mosaic”. I tend to think about self-worth as a mosaic that is the most resilient when you add different facets to it. So when you only fill your self-worth with one single thing (maybe only work, a partner or being an athlete) and that “shatters”, it’s devastating. But when your “you” is made up of multiple identities, it’s not the entire self-worth shattering from a setback in one smaller area of life. So mosaic it is.
I have also made a Bingo card with goals for the year with some bigger and smaller things. This includes taking a hand-lettering and a glass-blowing workshop because it just seems really cool. But I am also setting the goal of hitting 10k subs on YouTube (almost double), seeing the Northern Lights, and donating blood.
Can’t wait to see how motivating that will be in different areas!
Oh, and I’ll also set myself 30-day challenges again because they’re usually really fun! Starting out with exercise this month.
Thing I Like
With my 30-day exercise challenge underway, I love my thin Yoga mat. It just feels like the “exercise space” now, which is super helpful!
Hot off the Press
I subscribed to Gergely Orosz, Pragmatic Software Engineer, and he has an interesting ethics statement published.
In the past two years, I have never taken sponsorship for anything published in this newsletter and was considering how I could make this clear.
I ended up writing my own Ethics statement about my content online to clarify how I am influenced by existing requirements and financial incentives. It’s way less interesting than Gergely’s, but I think it gets the point across. Especially in today’s “influencer world”, with AI sponsorships and scams flooding my inbox…
I also bought the domain late.email, so if you want to say something nice in a sentence or two about this newsletter and be featured on the landing page I’m building, please feel free to just reply to this email. Also, mention if you want your full name featured (and a link to your main social). It’s in the making, so I still have to figure these things out!
Python Deadlines
We’ve had a bunch of Python deadlines over the holidays!
PyCascades and PyCon Slovakia have already ended.
PyCon Italia is ending tomorrow!
PyCon Germany and Pydata Berlin “look” like they ended but had an extension over the holidays!
PyCon Lithuania and PyTexas will also close soon!
Machine Learning Insights
2023 was the year of generative AI.
We have seen Stable Diffusion, chatGPT with GPT-4, Dall-E 3 and many more advances.
I have published one of my biggest courses yet about prompt engineering for creatives and published a book about chatGPT for Creatives and Content Creators, which has found wide appeal.
2023 has also seen the rise of “long-termism” around “AI alignment”. The questionable narrative that we should secure our world against a possible future conscious, superhuman, malignant AI instead of mitigating existing threats and security issues coming with large language models. I wrote about this for the Software Sustainability Institute.
But it was also the year of AI-driven weather forecasts.
I was lucky enough to work for the ECMWF, which had the wise anticipation to hire machine learning experts before the boom. So, I contributed to the model that is now known as AIFS and helped operationalise multiple data-driven weather models from big tech companies such as Nvidia, Huawei and Google Deepmind. I still have nightmares about Jax.
The ECMWF MOOC about machine learning for weather and climate also reached thousands of students, with over 1,000 finishing and gaining a certificate!
Some personal highlights also include a guest lecture at Brown University and publishing ML.recipes, a resource for scientists who want to apply machine learning responsibly in their respective fields. This newsletter also reached more than 1,000 subscribers, which is incredible, so thank you for such a lovely year!
Data Stories
2023 was the year of generative AI, and McKinsey agrees.
Between copyright lawsuits and new, bigger, and better models every day, tools became available to the public.
OpenAI made over a billion dollars in revenue because of chatGPT Plus.
And whether we wanted to or not, people became acutely aware of what modern compute systems are capable of.
McKinsey recaps the year through different charts.
This one about industries changing due to AI and automation was quite interesting to pick out.
Automating boring emails in the office and legal professions has catapulted their automation potential by a massive margin.
Source: McKinsey
Question of the Week
- How do you ensure the robustness of machine learning models against adversarial attacks?
Post them on Mastodon and Tag me. I’d love to see what you come up with. Then I can include them in the next issue!
Tidbits from the Web
- In case you need some rest after 2023, watch this
- How to become a full-time content creator as a software engineer
- This 13-year-old kid is the first to ever completely beat Tetris
Jesper Dramsch is the creator of PythonDeadlin.es, ML.recipes, data-science-gui.de and the Latent Space Community.
I laid out my ethics including my stance on sponsorships, in case you're interested!