Â鶹´«Ã½

Feedback is Â鶹´«Ã½â€™s popular sideways look at the latest science and technology news. You can submit items you believe may amuse readers to Feedback by emailing feedback@newscientist.com

Goblin mode

When you have been around as long as Feedback, you run the risk of falling into a rut. Repeating yourself, recycling the same old themes.

For instance, science-fiction author Charlie Jane Anders alerted us to the existence of a marine biologist called , but on careful inspection of Feedback’s well-ordered archive, we discovered we had mentioned him three times already, in 2005, 2013 and 2016. We can’t do the same case of nominative determinism for a fourth time, we decided: we’ll give the poor guy a complex.

The same problem of getting fixated on a concept appears to have befallen ChatGPT, the AI chatbot created by new . , the company revealed a afflicting its models: “they increasingly mentioned goblins, gremlins, and other creatures in their metaphors”.

The problem seems to have arisen because the company created a “Nerdy” personality option for ChatGPT (to go along with “Default”, “Cynical”, “Candid” and various others). This one was trained to be “playful” and to “undercut pretension”.

OpenAI says that, while training Nerdy, the company “unknowingly gave particularly high rewards for metaphors with creatures”, such that “Nerdy accounted for only 2.5% of all ChatGPT responses, but 66.7% of all ‘goblin’ mentions in ChatGPT responses”.

The firm eventually had to nip it in the bud, partly by retiring Nerdy and partly by an instruction to a newer version of ChatGPT: “Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query.”

Feedback notes, in the interest of due diligence, that this entire story is based on the official account given by the company. A cynic might say that OpenAI openly tells us fun stories like this to distract from the other things it is doing, but Feedback retired our Cynical personality, so we won’t.

As the various chatbots settle in, they do seem to be developing their own odd quirks. Reader Jo Edkins writes in to bring us up to speed with Anthropic’s AI Claude, which readers may recall was previously given the task of running a vending machine in the company’s offices, only to make a hash of it (26 July 2025). In December 2025, the company announced it had , this time using a pair of AIs called Seymour Cash (the boss) and Claudius (the employee).

The pair would have peculiar conversations. “We’d sometimes wake up to find that Claudius and Cash had been dreamily chatting all night, with conversations spiralling off into discussions about ‘eternal transcendence’, ” Anthropic reported.

Jo says she asked her son, “who has some knowledge of the field”, about chatbot-chatbot conversations. He told her: “Anthropic’s models in particular have a tendency to end up talking about cosmic bliss if you leave them talking to each other for too long. No one really knows why.”

This all makes Feedback feel better about our Haddock fixation.

It’s in the wind

Occasionally, we wonder what we would do if we won the lottery. Boring stuff like paying off our mortgage, of course, but then maybe we would go see a solar eclipse or visit Svalbard.

US President Donald Trump has taken a different approach. Finding himself with money burning a hole in his pocket, he has dropped nearly $2 billion to make sure offshore wind farms don’t get built. In March, he agreed to to French energy giant TotalEnergies not to build wind farms off the coasts of North Carolina and New York, and in April, to Bluepoint Wind and Golden State Wind not to build wind farms off New Jersey, New York and California.

The costs of net zero may be an issue, but who knew not doing it would be so expensive?

Ticking along

Bob Naylor headed his email to Feedback, simply, “Really???” This is always a good sign.

Bob had seen a post on a local forum promising “Natural Tick Protection for You & Your Dog”. It recommended “EM® beads”, which are “made from EM® ceramics – clay fermented with Effective Micro-organisms®, then baked to create tiny beads that emit far infrared waves”. These waves – stay with us here – “penetrate the skin up to 7.5cm deep, helping improve circulation and maintain a healthy microbiome – naturally deterring ticks along the way!”

Feedback’s first thought was that this was the Avengers: Doomsday of fruitloopery, since it features so many of the tropes. Next, we thought that 7.5 centimetres is really quite deep, unless you are truly thick-skinned. Scouting the internet, we found a number of sites selling dog collars with these beads.

Falling further down the rabbit hole, we learned that Effective Micro-organisms^® were the brainchild of a researcher called Teruo Higa in the 1980s and 90s. His original plan was to use them for things like cleaning septic tanks, but he ran into problems reproducing his results.

Feedback will go out on a limb and suggest that, even if the Effective Micro-organisms^® do work for their original intended purposes, we doubt that baking them at the temperatures required to produce ceramic beads would turn them into sources of invisible, anti-tick radiation.

Got a story for Feedback?

You can send stories to Feedback by email at feedback@newscientist.com. Please include your home address. This week’s and past Feedbacks can be seen on our website.