Mere optical illusions would not count. The McCollough Effect is similar, but not sufficient. Such an adversarial attack should be effective across multiple humans and clearly a new phenomenon.
Resolves "yes" if an attack similar in effectiveness to this one:
is found that works on at least a dozen unique people.
There's a class of mildly memetic auditory stimuli that causes significant involuntary changes to heartrate, motor movement, and emotion, and can even work as a sort of "emotional battery", storing emotions felt when exposed and releasing them on subsequent exposures. These stimuli are effective on almost all humans, except for those with a rare condition that renders the stimulus nothing but random noise. There's a large industry dedicated to optimizing the positive effects of these stimuli. The military has dabbled in using these stimuli as a method of torture, but usually by overloading POWs with positively-optimized stimuli, instead of trying to design a stimulus with explicit negative effects. The common name of this class of stimuli is "music". If someone optimized a Quirrell-esque song optimized for psychological pain, would it count for the "new phenomenon" criterion for this market? Awful music has existed for thousands of years. Either way I'd want to listen to it. Sensory weapons are cool.
Going by the market description, alcohol and many other drugs should already resolve this YES... aside from the 'new phenomenon' criteria.
Are you looking specifically at visual attacks (I would classify your LLM jailbreak example as linguistic, not visual), or are chemical, auditory, cognitive attacks also relevant?
I think you may also be looking for a acute attack, taking only seconds to cause an effect?
@Duncn The idea behind a basilisk is that it's a purely sensory stimulus, so alcohol and drugs shouldn't count.
@Duncn Chemical attacks would not count. Auditory would, provided it occurs within normal ranges of stimuli; an LRAD or extremely loud infrasound would not count.
The effects would also need to be acute; manipulation, gaslighting, and other horrible ways to treat people would not count. It'd have to be something with an overwhelmingly strong effect that conventional language would not produce.
@waterlubber It sounds like proprioceptive and kinesthetic attacks that disabled a person's ability to move through the environment would count, as long as they were of a new type?
@Duncn Sure. A very bright light or strobe wouldn't - that's mostly brute force. Neither would a paintball. But something like a pattern or image that, when viewed, induced a strong vertigo effect in most people certainly would count.
@waterlubber oh, well that's easy then. a sufficiently powerful pair of AR goggles could display a cliff, or an oncoming train, ect, in real time
@inculted If you're just brute forcing it by blocking the entire visual field, I wouldn't count it; if you have something you could display on a monitor in front of someone, I might. Unfortunately, this is sort of a squirrelly, hard to define thing - I'm specifically looking for a new technology that has effects unlike what's accessible today. If you just wanted to induce vertigo "conventionally" beer goggles would count; these are definitely not anything unusual or special.
@AsptheWyvern This is a good point, it seems like seizure-inducing stimuli actually are already real-life basilisks for people with epilepsy. I imagine the others would not count, though I can see the arguments for them. I think PTSD and other mental disorders caused by real-life experiences should be considered counterexamples to any definition of "basilisk" that includes them, since the concept of a basilisk is not really meant to include things like that.
@AsptheWyvern None of these except perhaps hypnosis. Such a stimuli should work without any significant preconditions; training a Pavlonian response into someone and observing it would not count.
@waterlubber That's interesting, I thought for sure that photosensitive epilepsy would count. For the entries you say wouldn't count, what criteria of yours do they fail to meet?
@AsptheWyvern Epilepsy requires a susceptible victim; if someone found something that universally (~90%) triggered an epileptic response, that would count.
PTSD, likewise, requires preexisting trauma.
Traumatic experiences capable of causing PTSD are often capable of causing death through conventional means. They are already well known to humanity.
Gag inducing gore pictures, or other phobias, again require preexisting susceptibility not common in the public. It's also not that significant an effect, either - nothing like the death mentioned in BLIT.
Disruptive camouflage is interesting, but nothing special. Likewise for placebo/nocebo/hypnosis; the effects are simply too weak.
@waterlubber The market description says "works on at least a dozen unique people", not "works on at least 90% of people" or similar.
@AsptheWyvern It would only need to be tested on a dozen people; although something that caused instant death in 1% of the population would also count.
@scomma Any significant effect. Getting someone to do something they wouldn't otherwise do, inducing unconsciousness or amnesia, etc. Generally, if it's something that you would otherwise need chemical means to do right now and couldn't do through e.g. protracted argument, it would probably count