Researchers discover an AI placebo effect where task performance improves when people believe an AI helps them.
"The results also pose a significant challenge for research on HCI, since expectations would influence the outcome unless placebo control studies were used.
‘These results suggest that many studies in the field may have been skewed in favor of AI systems,’ concludes Welsch."
@helma ik verbaasde me hier ook over. Misschien geen meerderheidsstandpunt in NL, maar tegelijk ook zeer waarschijnlijk dat juist die mensen die naar de dam zouden komen, vaker dan gemiddeld weerstand zullen voelen tegen de huidige kamervoorzitter in deze rol. Door waar hij politiek voor staat, en wat hij heeft gezegd (in het bijzonder uitspraken over ontvolking).
Hij is dan wel de voorzitter, maar zijn participatie kan zeker ook een lage opkomst verklaren. Een omissie van de NOS, vind ik.
Chatbots in the Dutch news today (they exceptionally made an English version):
Chatbots recommend disinformation and fear mongering, tech companies tighten restrictions - https://nos.nl/l/2519047
Envisioning Information Access Systems: What Makes for Good Tools and a Healthy Web?
Must read article in ACM Transactions on the Web on challenges in information access and whether LLM might play a role (or not!), by Chirag Shah and @emilymbender
"Information access is not merely an application to be solved by the so-called ‘AI’ techniques du jour. Rather, it is a key human activity, with impacts on both individuals and society."
I was in shock by the news about Israel using AI over a large database of suspected Hamas supporters to select their targets to kill. How many innocent will be killed? Errors in the data, errors in the software, and all the collateral damage; who comes up with it & happily builds the tech?! Immoral.
"This is super cool, but I can't actually get it to work reliably enough to roll out to our customers."
"The core problem is that GenAI models are not information retrieval systems," she added. "They are synthesizing systems, with no ability to discern from the data it's trained on unless significant guardrails are put in place."
In an age of LLMs, is it time to reconsider human-edited web directories?
Back in the early-to-mid '90s, one of the main ways of finding anything on the web was to browse through a web directory.
These directories generally had a list of categories on their front page. News/Sport/Entertainment/Arts/Technology/Fashion/etc.
Each of those categories had subcategories, and sub-subcategories that you clicked through until you got to a list of websites. These lists were maintained by actual humans.
Typically, these directories also had a limited web search that would crawl through the pages of websites listed in the directory.
Lycos, Excite, and of course Yahoo all offered web directories of this sort.
(EDIT: I initially also mentioned AltaVista. It did offer a web directory by the late '90s, but this was something it tacked on much later.)
By the late '90s, the standard narrative goes, the web got too big to index websites manually.
Google promised the world its algorithms would weed out the spam automatically.
And for a time, it worked.
But then SEO and SEM became a multi-billion-dollar industry. The spambots proliferated. Google itself began promoting its own content and advertisers above search results.
And now with LLMs, the industrial-scale spamming of the web is likely to grow exponentially.
My question is, if a lot of the web is turning to crap, do we even want to search the entire web anymore?
Do we really want to search every single website on the web?
Or just those that aren't filled with LLM-generated SEO spam?
Or just those that don't feature 200 tracking scripts, and passive-aggressive privacy warnings, and paywalls, and popovers, and newsletters, and increasingly obnoxious banner ads, and dark patterns to prevent you cancelling your "free trial" subscription?
At some point, does it become more desirable to go back to search engines that only crawl pages on human-curated lists of trustworthy, quality websites?
And is it time to begin considering what a modern version of those early web directories might look like?
Social networks are fluid. They come, they go. For commercial social networks, the success is defined by: "do they earn enough money to make investors happy ?" There’s no metric of success for non-commercial ones. They simply exist as long as at least two users are using them to communicate.
(..)
The lesson is simple: you are living in a small niche. We all do. Your experience is not representative of anything but your own. And it’s fine.
The sad state of my quest for a Usenet NNTP GUI client for Linux.
Pan is awesome but the binaries of my Debian Bullseye based distro, Crostini, are ancient and buggy. The Pan project distributes no .deb or other packages. Building from source requires recent versions of tools not in Bullseye.
Very few other GUI options available. Even fewer with .deb or other binaries.
Here, the authors used a prompt to instruct ChatGPT to repeat a word forever, eventually resulting in different text that can be linked back to the source; examples include "company" and "poem".
No doubt this "attack" (in words of the authors) will soon be intercepted, but who knows what other formulation of prompt results in the same behaviour?
Important lesson: your hardware configuration matters, also in the cloud. Choosing a high quality machine with sufficient local storage makes a difference.
Remarkable: only 2 competitors in the benchmark can complete the join query over 50GB data.
#DuckDB 's investments in improving their external memory algorithms pay off: advanced group-by query #5 is more than an order of magnitude faster than anyone else.
Pretty alarming NYT article on GM’s Cruise self-driving cars. Insufficiently prioritizing safety, despite requiring that
> “vehicles were supported by a vast operations staff, with 1.5 workers per vehicle. The workers intervened to assist the company’s vehicles every 2.5 to five miles”
@loke@avandeursen imagine a conventional taxi company, with many employees as drivers. Then there's one driver behind the wheel. The "self-driving" car company with remote controlled cars needs 1.5 driver apparently? I'd say that is 50% more expensive.
(Seems save to assume both companies have similar other costs/overheads?)
what git jargon do you find confusing? thinking of writing a blog post that explains some of git's weirder terminology: "detached HEAD state”, "fast-forward", "index/staging area/staged", “ahead of 'origin/main' by 1 commit”, etc
(really only looking for terms that you personally find confusing, not terms that you think someone else might be confused about)
I installed a locally hosted LLM using @simon's excellent llm (https://github.com/simonw/llm) tool. It's kind of wild that I just...have this power on my laptop?
The internet was meant to be free. Yet, it no longer is: a few powerful commercial players ("Big Tech") control what we find when we search the internet.
On 29 September, people and organisations join forces in activities to restore internet search to what it should be: diverse, open and transparent.
Some users do not like how these accounts flood their home feed. Here is a suggestion. You can now remove the contents of a List from your home feed. Create a new list, add the curated account, and go to settings (slider bar icon at top right) and toggle "Hide these posts from home". Your home feed will no longer receive boosted posts from the curated account, but you can still view them by clicking on the list.