A man looks in his bathroom mirror in the morning, points at the reflection with surprise and recognition. The caption reads: "I'll know I'm famous when I look in the mirror one morning and say, "Hey! I *know* that guy!"
Look ma. I made a meme!
A man looks in his bathroom mirror in the morning, points at the reflection with surprise and recognition. The caption reads: "I'll know I'm famous when I look in the mirror one morning and say, "Hey! I *know* that guy!"
Look ma. I made a meme!
www.mdpi.com/2073-431X/12... is a good starting point. (Read it all.) In essence, a neural net knows only tokens. It has no way evaluate data, to know if a sequence of tokens represents a fact, fantasy, bias, or delusion. It has no way to detect missing data. False output is unavoidable.
Some of that is true, but the problem has nothing to do with recursive training: It's inherent in a neural net trained on textual data. The fundamental structure of an LLM means it can't be made to "never hallucinate," will get things wrong from time to time, and has no way to validate itself.
What is seldom mentioned is that this is not a bug that can be fixed, any more than a pig's inability to fly can be "fixed." It's inherent in the architecture.
One did not need to be a computer scientist to know this. The fundamental architecture of LLMs ensures they can't be a path to superintelligence, and guarantees that they will make stuff up (hallucinate).
I think many of us knew this would happen. While it would be wonderful to imagine that it's because Congress wants to invest in NASA, the sad truth it has far more to do with pork barrels than science.
The latest Cross-Product Creative Challenge is live on the Adobe Firefly community forum. I can neither confirm nor deny that may have slipped in a little political satire on this one.
community.adobe.com/t5/adobe-fir...
How is this news? We've known for at least a decade that faster charging degrades batteries more rapidly. Was there some idea that this didn't apply to EVs?
Look. It's the Daily Mail. That lot wouldn't know a fact if it hit them in the face. Think "National Enquirer" with added sleaze.
In Summer, anyway. Australia is at 1/3rd of homes on solar plus battery, an astonishing statistic in the face of rampant coal lobbying and climate denialism.
Please don't wreck Bluesky with political garbage the way MAGA wrecked Twitter. Enough, already!
An important discussion about the impact of AI on creative professions by two people at the leading edge of the field: Mark Heaps and Amy Balliet. This is for anyone who's a designer, photographer, or artist. #creativity #ai #design
www.youtube.com/watch?v=6Igd...
If you're a creative, working with Adobe tools, feel free to join in my monthly Cross-Product Creative Challenge on the Adobe Firefly forum.
community.adobe.com/t5/adobe-fir...
My friend Amybeth Menendez, a consummate professional in the world of print design, posted an excellent conversation starter on the Adobe Community forum. How do we integrate AI into a professional workflow where craft and detail matter most?
community.adobe.com/t5/photoshop...
Iβve spent a good chunk of my career relying on American science and engineering to keep me alive. Yesterday, RFK Jr. testified at the U.S. Senate Finance Committee. It was sad to see him try to destroy the lifeβs work of so many American scientists. He shouldnβt be in this job.
I had no idea there was such a thing as "The Journal of Internet Medical Research." It really exists, and a new research paper on using short video to reduce vaccine hesitancy looks well done (and a good idea).
www.jmir.org/2025/1/e66758
More than a little misleading, especially with the GenAI image.
Magellan never said that. As far as I can tell, this was from an 1873 essay by an American evangelist of agnosticism called Robert Ingersoll. Belief in a flat Earth was never a dogma of any European or Middle Eastern church.
Here's the thing about AI-generated marketing:
That little shimmy a 0:02 is cute. π
RFK Jr. Falsely Claims Measles Vax Causes Deaths 'Every Year' www.medpagetoday.com/infectiousdi...
Poster image for the Print Design Summit
Print Design Summit coming up! Free tickets at this link: www.printdesignacademy.com/summit
Urban night scene. Menacing robot in foreground. Quote text: "The whole aim of practical politics is to keep the populace alarmed (and hence clamorous to be led to safety) by menacing it with an endless series of hobgoblins, all of them imaginary." β H. L. Mencken
Consider the state of a society in which a lifelong dedication to saving lives is rewarded with political persecution and personal danger:
www.medpagetoday.com/special-repo...
Anyone who uses or is interested in #AI should view the recent #AIforScience forum from the Royal Society and Google DeepMind. Playlist here: www.youtube.com/watch?v=mGsy...
Post from Jerome Adams on X showing statistical evidence that vaccination reduced COVID-19 deaths. 37.4% COVID deaths were unvaccinated, 0.8 percent deaths among fully vaccinated.
For your friends who are confused about vaccines. Florida surgeon-general take note.
#vaccines #science
An #AI prompted image is only the start. Use #Photoshop or #Illustrator and breathe life into it with photos, brushes, drawings, genFill, genExpand. A composite with your creativity in it is more alive, more interesting, and more satisfying than just generated pixels. #AIart
I use ChatGPT to help me write tutorials.
It is often WRONG. I think itβs a fantastic tool for subject matter experts to use who can distinguish whatβs true and make corrections but donβt rely on it for the truth.
www.theverge.com/2024/12/5/24...
Google's NotebookLM is evolving into an outstanding research tool. AND it's fun to play with. The "podcast" feature is getting the ability to take feedback from the director (you) and modify the "hosts'" conversation.
www.youtube.com/watch?v=mccQ...
That is hilarious! π