🐍 Python 3.14 releases this year, making it π-thon
In this issue, we talk about AI in wildlife conservation. The anti-woke tech bro. Hacking LLMs with invisible text and running AI weather models on open data! I also made some cute mascots for Pythondeadlines, which I’m proud of!
Late to the Party 🎉 is about insights into real-world AI without the hype.
Hello internet,
It’s 2025. Happy new year! 🎉
(We still have over 350 days that can get more positive…)
In this issue, we talk about AI in wildlife conservation. The anti-woke tech bro. Hacking LLMs with invisible text and running AI weather models on open data! I also made some cute mascots for Pythondeadlines, which I’m proud of!
Let’s dive right in and hack some machine learning!
The Latest Fashion
- More creative vulnerabilities: Hacking LLMs with invisible text
- The cultural significance of the “anti-woke” tech bro
- You can now run AI weather models completely from ECMWF open data
Worried these links might be sponsored? Fret no more. They’re all organic, as per my ethics.
My Current Obsession
I've been watching the political pendulum swing with a knot in my stomach. There’s a clear shift and it has been happening for the last decade at least, but everyone looks a bit funny when I talk about the signs. So there’s a clear feeling of anxiety about the global political climate. The middle ground - once promised a place of reasonable compromise - seems to be eroding beneath our feet; in Germany, even the Green Party is talking about deportations now, apparently. I find myself lying awake some nights wondering about my friends, about what these shifts mean for our marginalised communities, but for all of us really. Because as we know it won’t stop with women’s reproductive rights or policing away trans people.
But then, sometimes, the world serves up moments of pure irony that make you laugh despite everything. Have you been following the great American migration to Red Note / XiaoHongShu? There's something delightfully absurd about watching folks practising their Mandarin when usually they just talk English louder to make you understand. Just waiting for the "Nǐ hǎo, y'all!" as the new cultural fusion we didn't know we needed. The internet always finds a way.
Speaking of digital spaces, I've been having entirely too much fun creating these little Python mascots for PythonDeadlin.es. Sometimes when the world feels heavy, you need to make something purely joyful, even if it's just a goofy reptile derping on my website.
And for personal updates, I have started losing weight with the help of weight loss medication (and, of course, medical supervision, I’m not that reckless). I’m down over 25kg (55lbs). I have also started working out, and as someone who has done a lot of “lonely weightlifting in gyms”, I have now discovered the joy of workout classes. There’s just something about suffering together that is really nice when working out. The fact that I can now chase after trains and buses without problems and hold conversations going upstairs (on good days, let’s be real) is such an improvement in my well-being. Especially seeing what a toll the last year ever since my PhD has taken on me. If you’re in a larger city in Europe, check out Urban Sports Club, it’s a gympass type of thing, where you can go to courses in multiple participating gyms, even the fancy boutique ones in my area. Really liking it.
Thing I Like
I’m currently trying the Bose Quiet Comfort Ultra after another pair of Sony started malfunctioning. So far, I really like them, but what you read online is true. The sound on the Sonys is unsurpassed. The noise cancellation and interface of the Bose are pretty great, though. So far, I think I will keep them.
Hot off the Press
I made a Python pun, which went semi-viral, and of course, I recycled it here! I also talked about the new snake icons I added to PythonDeadlin.es which got their proper attention!
On Linkedin, I talked about improving ensemble weather forecasts, which was shared around a bit.
I announced I would be deleting my Meta accounts on the weekend as a consequence of Meta’s lobbying against TikTok and general right-wing pandering.
Python Deadlines
I found these conferences PyCon UK and PyCamp Argentina and new CFPs for Scipy US, Pyconf Hyderabad, and EuroPython.
And for upcoming deadlines, currently, there is only EuroPython.
Machine Learning Insights
Last issue I asked, What are the ethical implications of using AI in wildlife conservation efforts?, and here’s the gist of it:
The implementation of AI in conservation introduces complex privacy considerations, particularly regarding indigenous communities and local populations. While satellite monitoring and drone surveillance offer powerful tools for tracking wildlife and detecting illegal activities, they simultaneously raise concerns about surveillance overreach and potential privacy violations of communities living in or near protected areas.
As we saw in the deforestation monitoring example, these technologies can inadvertently capture data about the lawful activities of Indigenous peoples, potentially leading to misidentification and unnecessary confrontations. This presents a fundamental tension between conservation goals and human rights.
Equity and Access
The distribution of AI conservation technologies reveals significant disparities in implementation capability. Many regions facing the most pressing conservation challenges often lack:
- Adequate computational infrastructure
- Reliable internet connectivity
- Technical expertise for system maintenance
- Financial resources for high-resolution data acquisition
This creates an ethical imperative to address these inequities while developing solutions that work within existing constraints.
Practical Implementation Challenges
Resource Allocation
The deployment of AI conservation systems requires substantial investment in:
- Hardware infrastructure
- Data collection systems
- Training programs for local personnel
- Ongoing maintenance and updates
This raises questions about opportunity costs: Could these resources be better allocated to other conservation approaches? When working with limited conservation budgets, we must carefully weigh the benefits of AI implementation against other potential interventions.
Accuracy and Accountability
AI systems can produce false positives or negatives in wildlife monitoring, potentially leading to:
- Misdirected conservation efforts
- Unnecessary interference with wildlife
- Inefficient use of limited resources
Who bears responsibility when AI systems make mistakes? This question becomes particularly crucial when decisions affect endangered species or impact local communities' livelihoods.
Future Considerations
Long-term Sustainability
We must consider the long-term implications of AI dependency in conservation efforts:
- Will communities become overly reliant on these technologies?
- How can we ensure knowledge transfer and local capacity building?
- What happens if funding for AI systems becomes unavailable?
Power Dynamics
The implementation of AI in conservation efforts often involves multiple stakeholders with varying degrees of power and influence. We must carefully consider:
- Who controls the technology?
- Who has access to the data?
- Who makes decisions based on AI insights?
- How can we ensure equitable participation in decision-making processes?
Recommendations
To address these ethical challenges, I propose several key considerations:
- Prioritize inclusive design processes that incorporate local knowledge and perspectives from the outset
- Develop clear protocols for data governance and privacy protection
- Invest in local capacity building to reduce technology dependence
- Create transparent accountability frameworks for AI-driven decisions
- Establish mechanisms for regular ethical review and adjustment of AI systems
Conclusion
The ethical implications of AI in wildlife conservation extend far beyond technical considerations. While AI offers powerful tools for conservation, its implementation must be guided by careful consideration of equity, privacy, and long-term sustainability. Success requires balancing technological capabilities with social responsibility and environmental stewardship.
These ethical considerations should not paralyze conservation efforts but rather inform more thoughtful and inclusive approaches to AI deployment in wildlife protection. Through careful attention to these issues, we can work toward conservation solutions that benefit both wildlife and human communities.
I believe it's crucial to continue examining these ethical dimensions as AI technology evolves, ensuring that our conservation efforts remain both effective and ethically sound. What are your thoughts on balancing these competing priorities in conservation technology?
Got this from a friend? Subscribe here!
Question of the Week
- What are the key considerations in using NLP models for sentiment analysis in diverse languages, and how do you address cultural nuances?
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
Now this one’s a funny one. I used to share Tiktoks a lot here lately. And I won’t start sharing Reels any time soon. Let’s see how this category evolves.
- Here’s one of the many reasons I love Dimension 20 so much.
- Not a video?! How flash games shaped the history of the internet.
- How 2025 is going…
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!