🧵 Don’t let the Thread-bugs bite
Late to the Party 🎉 is about insights into real-world AI without the hype.
Hello internet,
I wasn’t doing super well this week, but somehow things keep happening. I have some really fun machine learning (and other content) today. So let’s dive in!
The Latest Fashion
- The Introduction to Statistical Learning now comes in Python!!
- I love this collection of papers and tech blogs in data and ML
- Google announced the first Machine Unlearning challenge
Got this from a friend? Subscribe here!
My Current Obsession
My Youtube now has 5,000 subscribers! 🎉 Quite stoked about that and trying to make more useful videos for a broad audience of tech nerds and ML enthusiasts.
Twitter is dying, and Facebook just published a competitor that may be the coup-de-grace. It has been unhinged and lovely. I have a collection of machine learning people looking for their kind right here! It feels a lot like old Twitter, and the vibes are immaculate. My feed is mostly made up of Chris Albon’s posts. Nothing to complain about there. (And they’re waiting for you… I asked.)
I will publish my latest Skillshare course, “Unlock your Creative Potential with AI: ChatGPT for Content Creators”, after this weekend! I have also created an ebook that goes with the course. It should also stand alone, and I’m incredibly proud of both. I hope to cut through the noise of subpar advice on chatGPT and give creators a solid understanding of this monster.
I made the decision to start the migration of the Latent Space to Discord. I got the advice to use Discord before. I should’ve listened. I got more and more messages that emails from the latent space weren’t coming through, which is unacceptable. I’m still setting things up over on Discord, but you can already join over at latent.club
Thing I Like
My Ninja air fryer has been putting in work these last few days. And I think it’s on sale in America right now.
Hot off the Press
I uploaded a short about missing data in machine learning!
In Case You Missed It
More people are finding ML.recipes, which is awesome!
Machine Learning Insights
Last week I asked, How would you communicate a machine learning solution to subject matter experts? and here’s the gist of it:
Effectively communicating a machine learning solution to subject matter experts is crucial for ensuring understanding, collaboration, and successful implementation. Here are some steps to communicate a machine learning solution to subject matter experts:
- Understand the Audience: Begin by understanding the background and expertise of the subject matter experts. This will help you tailor your communication to their level of understanding. Determine their familiarity with machine learning concepts and adjust your explanations accordingly.
- Provide Context: Start by providing a high-level overview of the problem being addressed and why a machine learning solution is being considered. Explain how the solution can provide valuable insights or enhance decision-making processes. Use relatable examples from their domain, such as meteorology, to illustrate the potential impact of the machine learning solution.
For instance, when communicating a machine learning solution for weather forecasting, you could emphasize how it can improve accuracy, help identify weather patterns, or provide early warnings for severe weather events.
- Explain the Approach: Clearly describe the machine learning approach, avoiding technical jargon. Explain the types of data being utilized, such as meteorological data, and highlight the machine learning algorithms or techniques employed. Focus on the benefits and strengths of the chosen approach and how it relates to the subject matter experts’ domain.
- Demonstrate Results: Present tangible results and outcomes achieved through the machine learning solution. Use visualizations, graphs, or charts to illustrate the improvements or insights gained. Emphasize the relevance of the results to the subject matter experts’ work and how it can support their decision-making processes.
For example, you could showcase how the machine learning solution accurately predicted severe weather events or provided long-range forecasts with high precision, enabling better preparedness and response planning.
- Address Limitations and Interpretability: Acknowledge the limitations of the machine learning solution, such as uncertainties or potential sources of error. Explain how these limitations are managed and discuss any interpretability challenges associated with the model’s decision-making process. Offer insights into how the subject matter experts can interpret and trust the results generated by the model.
- Encourage Collaboration and Feedback: Foster an environment of collaboration and invite subject matter experts to provide feedback, insights, and domain-specific knowledge. Engage in a dialogue to ensure their perspectives are considered and identify potential improvement areas or further customization of the machine learning solution.
- Provide Support and Training: Offer assistance and training to subject matter experts, particularly in understanding the outputs of the machine learning solution. Provide resources, workshops, or documentation to help them effectively interpret and utilize the results. Address any questions or concerns they may have and provide ongoing support as needed.
Data Stories
Twitter is dying.
We know that even if we don’t want it to be true. But a company without staff that doesn’t pay its bills is not exactly the poster child of corporate health.
And then Threads came along. People gave Mastodon a try. They flocked to the Bluesky waitlist. Then Facebook said, “hey, we made a Twitter clone”, and the social internet exploded.
People measured the immense chatGPT success by “time to 1 Million users”, and it looks like that one was below days with Threads.
It’s an unhinged space and quite fun right now. Find me as “jesperdramsch” as per usual.
Source: Statista
Question of the Week
- Can you write an end-to-end machine learning model in 500 characters?
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
- They challenged chatGPT to create recipes out of random ingredients and it’s actually good?
- How to find the will to live and Komorebi
- John Green on the seduction of despair (I love the long-cord aesthetic…)
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!