[scq]We created a robot tour guide using Spot integrated with Chat GPT

A recently updated list of examples of how AI models have exploited loopholes in the task specification or environment to achieve their goals.

 “I know there are a million people sharing temperature anomaly charts and maps lately, but there's a good reason for that. This is totally bonkers and people who look at this stuff routinely can't believe their eyes. Something very weird is happening.

“Firefly is our creative generative AI engine. It’s just landed in Adobe Photoshop.

“Through DragGAN, anyone can deform an image with precise control over where pixels go, thus manipulating the pose, shape, expression, and layout of diverse categories such as animals, cars, humans, landscapes, etc. As these manipulations are performed on the learned generative image manifold of a GAN, they tend to produce realistic outputs even for challenging scenarios such as hallucinating occluded content and deforming shapes that consistently follow the object's rigidity. 

“Announcing GPT-4, a large multimodal model, with our best-ever results on capabilities and alignment.

“So I like the idea of being stingy with our goals for the new year. Make that New Year’s resolution to listen to fewer podcasts, to travel less, to never go camping. Pick things you explicitly won’t do. Your time is finite. A new year is not the time to pursue infinite goals; it’s a time to pick the important ones. And not everything can be important.

According to a 2021 study of 60 boxing and MMA fights, 78% of the fights led to a concussion, and the fighter that sustained the first concussion ultimately lost 98% of the time.

Seems like strike-based martial arts amount to "last one to get a concussion wins".

No good evidence that tailoring teaching to students’ preferred learning styles has any effect on objective measures of attainment. The effect of “nudges” (clever design of defaults) may be exaggerated in general. No good evidence of anything from the Stanford prison ‘experiment’. No good evidence from the famous Milgram experiments that 65% of people will inflict pain if ordered to. Lots of screen-time is not strongly associated with low wellbeing; it explains about as much of teen sadness as eating potatoes, 0.35%. The Pygmalion effect, that a teacher’s expectations about a student affects their performance, is at most small, temporary, and inconsistent, r<0.1 with a reset after weeks. At most weak evidence for stereotype threat suppressing girls’ maths scores. i.e. the interaction between gender and stereotyping. Good evidence against ego depletion, that willpower is limited in a muscle-like fashion. Questionable evidence for the “hungry judge” effect, of massively reduced acquittals (d=2) just before lunch. Questionable evidence for a tiny “depressive realism” effect, of increased predictive accuracy or decreased cognitive bias among the clinically depressed. No good evidence for multiple intelligences (in the sense of statistically independent components of cognition). Mixed evidence for a small marshmallow effect. No good evidence that brains contain one mind per hemisphere. At most extremely weak evidence that psychiatric hospitals (of the 1970s) could not detect sane patients in the absence of deception. Questionable evidence for the menstrual cycle version of the dual-mating-strategy hypothesis. No good evidence that humans are cognitively or emotionally limited to 150 relationships with other people (Dunbar’s number).

“The mixture of cold fresh water from the glaciers and warmer seawater drives ocean circulation near the glaciers and farther out in the ocean, meaning the melt has far-reaching implications. The Greenland ice sheet is also important for sea-level rise; Greenland ice holds enough water to raise sea levels by 20 feet (6 meters).

“We’ve trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.

Among the sample conversations included are topics like "help me write a short note to introduce myself to my neighbor", "this code is not working like i expect — how do i fix it?", "what is fermat’s little theorem?" and "how to protect my home from burglary".

“The Mystery Flesh Pit is an extremely elaborate fantasy worldbuilding exercise created by writer and illustrator Trevor Roberts.

Imagine a Permian Basin-sized eldritch organism hibernating under the soil of Texas just off of Abilene, that will end the world should it ever awake. And instead of having the discovery of such a thing drive America insane in a bout of cosmic horror, we decided to turn it into a National Park Service-operated tourist trap in the vein of the Carlsbad Caverns. So we wired its jaw open, stuck an elevator and a Burger King franchise down its gullet, and sold guided tours and camping trips into the living interior of the beastie. Oh, there's also an all-inclusive resort by its digestive system and an aphrodisiac spa in its amniotic sacs.

The joke, as it were, is this really incongruous mashup of Lovecraftian horror subject matter, but described in the breathlessly enthusiastic tone of a tourist brochure, and all illustrated by Robert's incredibly realistic parody of the NPS's informational panel art.

 I just learned that these creatures exist.

“eDiffi is a new generation of generative AI content creation tool that offers unprecedented text-to-image synthesis with instant style transfer and intuitive painting with words capabilities.