Data Scientist (#MachineLearning, #DataViz, #NLP). Speaker. Teacher. Editor/co-author of The Handbook of #DataScience and #AI. Get it here: https://xitter.cf/t.co/gT6K9yD5ih
Zurich, Switzerland
Joined July 2020
- Tweets1.3K
- Following690
- Followers992
- Likes3.2K
Pinned Tweet
Everyone wants to build a #MachineLearning model. Not everyone knows what to do with its outputs.
I've found predicting Customer Lifetime Value is a good example. So I wrote a lot about what you can actually do about this kind of info. Enjoy!
#DataScience #Commerce #Marketing
ALT Hand-drawn picture of a documents folder, upon which is written "Operation SPILL ALL" and a stamp reading "DECLASSIFIED."
Congrats on your Customer Lifetime Value prediction model — now what are you going to do with it? by @KatherineAMunro buff.ly/3qeTKe7
New FREE contest for writers of Young Adult Fantasy, Sci-Fi, and Horror — tinyurl.com/wdwcon10 — Judged by Agent Keir Alekseii of Azantian Literary, via @writingdayworkshops. Winners get an agent critique of their work.
Boring ideas are screaming to become big businesses.
→ Build for transitions, not trends.
→ Charge for outcomes, not optimism.
→ Solve obvious problems, not hypothetical ones.
Wise words for the entrepreneurs and wannabe #startup founders out there: tinyurl.com/4b9by33w
Hey #AI, why you do this? 😅
ALT An AI generated image of a man working at a laptop where the screen is displayed on the back of the laptop case.
#AI-driven products are dynamic and unpredictable-a major challenge for adoption and user satisfaction.
The best ones guide users through a seamless, intuitive, and trustworthy experience.
Love this piece on how the top companies are going about it: tinyurl.com/2y2wy7yj
Two very different attitudes to #AI on my Medium feed today 😆
ALT Screenshot showing two suggested articles on Medium. The first is titled "I'm a Professional Editor and These Phrases Tell Me You Used ChatGPT", the second is "How We Use AI To Publish 50+ Human-Like Articles Per Month."
"#OpenAI’s board saying “We need to raise more capital than we’d imagined” just after raising $6.6B at a valuation of $157B sounds alarmingly like a Ponzi scheme. Basically, “Trust us, we can maintain our lead, we just need a never-ending stream of infinite investment.”"
Oof. 😬
New research from @AnthropicAI finds you can #jailbreak an #LLM just by hitting it with variations of the prompt with different capitalization.
futurism.com/the-byte/easy-h…
ALT The dog in a room full of fire meme, with the caption tHis iS fiNe.
The excuses big tech are using in court to get away with stealing copyrighted works are almost funny... until you realise the precedent they're setting.
I tried to figure it all out. This is what I learned.
katherine-munro.com/p/how-bi…
#RAG is a valuable use case for #LLMs & a boon for the quality of informational chatbots. But if someone tells you that RAG “solves” the hallucination problem, they're telling you they don’t understand how any of this works.
Couldn't agree more, @Rasa_HQ tinyurl.com/mrvs97k5
Are you trying to push an #AI initiative from the bottom up? How can you transition from such an emergent strategy to a purposeful one, i.e. with buy-in from the very top?
Good question; interesting article from @VUXworld
tinyurl.com/ykmywp84
This "Grand Unified Theory of the #AI Hype Cycle" really is grand. Short, simple, funny, and so right.
tinyurl.com/y5t37s4u
If you need to understand the top risks when building #LLM-applications, this is great.
It won't necessarily help you solve all your problems, but it will give you something to be scared about. 😅
tinyurl.com/mr24pj95
ALT Screenshot of the Top 10 LLM risks described in the linked PDF.
How do #LLMs handle contradictory instructions? E.g. when given examples differ from the system prompt? How SHOULD they behave?
Here's a nice experiment: tinyurl.com/3e49j6n9
It focuses on multilingual prompts, but I also see risks when whole teams work on a bot together.
Ran out of room to credit the wonderful @ArtFishAI, for yet another great article!
"It turns out language models love being given language and not a bunch of text snippets."
Nice. Somehow I find that pretty satisfying. And if creating a better coding copilot means my younger colleagues finally learn to DOCUMENT THE DAMN CODE! I'm all for it 😅
This Post is from an account that no longer exists. Learn more
As I said in this article, news of #GenAI model providers acting shady seems to pop up weekly. Latest example: Apple, Nvidia and Anthropic using scraped YouTube videos to train their models. Who would have guessed? 🤔
tinyurl.com/4km3sh4d
"I’ll use a number of current lawsuits against Stability AI, OpenAI and Meta, among others, to illustrate some criteria and considerations which may be used to evaluate whether or not an activity constitutes fair use." by @KatherineAMunro buff.ly/4eTF34k
Agree w/ concerns that using sparkles ✨ to mean #AI causes inflated expectations & hides negative aspects of the technology. But this makes a good point: at least it's given us a clear, cross cultural way to highlight #GenAI content, so people can consume it more consciously ✨
A look at the rise in use of the sparkles emoji to market AI products as magic, as some say the imagery distracts from real-world issues the AI industry faces (@rachelmetz / Bloomberg)
bloomberg.com/news/newslette…
📫 Subscribe: techmeme.com/newsletter?from…
techmeme.com/240711/p1#a2407…