When prompted to provide the current top news headlines from specific outlets, ChatGPT returned non-news output 52–54% of the time (almost always in the form of an ‘I’m unable to’-style message). Bard did this 95% of the time.
For ChatGPT, just 8–10% of requests returned headlines that referred to top stories on the outlet’s homepage at the time. This means that when ChatGPT did return news-like output, the headlines provided did not refer to current top news stories most of the time.
Of the remaining requests, around one-third (30%) returned headlines that referred to real, existing stories from the news outlet in question but they were not among the latest top stories, either because they were old or because they were not at the top of the homepage.
Around 3% of outputs from ChatGPT contained headlines that referred to real stories that could only be found on the website of a different outlet. The misattribution (but not the story itself) could be considered a form of hallucination. A further 3% were so vague and ambiguous that they could not be matched to existing stories. These outputs could also be considered a form of hallucination.
The outputs from ChatGPT are heavily influenced by whether news websites have chosen to block it, and outputs from identical prompts can change over time for reasons that are not clear to users."
The big problem with the ubiquitous push for #AIAssistants is the idea that I would want to talk to them if they just were a bit better.
Sometimes I don't even want to talk to real humans, so why would I ever want a fake human to be my interface to things that I can perfectly read myself?
BARD App iOS. For those that did not know, the BARD app has some built in Siri commands, but you have to activate them before they will work. You can do things like start, stop, different sleep timers, set bookmarks, and jump backwards and forwards by time increments. there are also a few interactive commands. To activate any of them, go to the settings tab, swipe to Siri Settings, Double tap the now reading audio button. The commands are divided into 2 categories; Playback, and Navigate, just double tap the category you want. For each command, there will be an add to siri button. If you double tap the button, it will add it to Siri and bring up a screen for that command. The command you will say to Siri will be in the edit field, just double tap to change it if you want. Once things are set, just hit the done button. It will put you back in the BARD app. Its that easy. I like the interactive commands. As you swipe through the list, it will tell you if it is interactive. What that means is that once the command is given, Siri will give you a list of choices to choose from. For example, there is a sleep timer interactive command. If you set this up, then when you say the command, Siri will come back and ask you if you want to set a 15 minute, 30 minute, 45 minute, 60 minute timer, or end of chapter sleep timer. You just tell her what you want, and it is set. I like the interactive commands because you are given choices. These are probably something that I will not always use, but it will be good to have some set up for times they can come in handy. #iPhone#iOS#BARD#Siri#Shortcuts#Commands#HandsFree
Sorry, ich kann es nicht anders sagen, aber unsere Kinder werden mit einer KI-Shitshow aufwachsen. Und es liegt an jedem Einzelnen von uns, diese KI-Shitshow nicht so schlimm werden zu lassen, wie ich sie mir aktuell vorstelle. 🥶
yeah ... classic for google again: crappy marketing below expectations, omissions in statements and other techniques that are obviously misleading and annoying.
In other news, Google’s Bard AI gave me a genuinely useful answer to a question last night. I could’ve obtained the same information through conventional search engines, but I would have to do multiple searches, find the relevant bits of information in several different articles, and then collate the information to do what Bard did in about a second.
My early impression, which is subject to revision, is that Bard can be immediately useful for comparisons and summaries of information that is already available and trustable.
In other words, it depends on what sort of questions you ask. Get familiar with the limitations and capabilities of today’s Large Language Models and make use of them where appropriate. They can save you time and help you be more productive, personally and professionally. But you have to put in the work to learn how to ride this new bicycle.
I saw that Google's Bard can generate images now, so I tried my go-to test of "Generate an image of a bear riding a unicycle in a snow storm". It looks like it still isn't as good as OpenAI in the Bear/Unicycle/Snow test.
Seeing people praise #copilot for finally getting rid of hallucinations through simple RAG techniques of checking for reality in eg. citations. This moment where a lot of the trivial claims against #LLMs stopped being true, but the deeper harms of surveillance and information monopoly remained was inevitable and the chief danger of dismissing it as "fancy autocomplete." That is why I wrote this almost a year ago, as a warning of what comes next and what we can do about it: https://jon-e.net/surveillance-graphs/ #SurveillanceGraphs
That part of "bypassing much of the agency individual people might otherwise have to avoid them" is reflected in #Google integrating #Bard into android messenger. While academics worry about AI sentience and toy with them as objects of study, this is capitalism baby, and the information conglomerates came to kill each other and capture the market now that investors are calling due on stagnant ad money. It doesnt matter if the LLMs work. That's about as effective of an argument as saying politicians lie. Theyre just the pretense for a new modality of surveillance spanning individuals, groups, businesses, and governments.
There are actually two Nano models, according to their paper. Bard is based on (a fine-tuned version of) the Pro model, with an upcoming premium tier featuring the (yet unavailable) Ultra model size.
#Bard users have noticed a fixed pre-span conversation to each chat session. It would make sense to hide that away behind a dedicated (internal?) Bard Assistant API, rather than exposing the full Google Vertex AI API.
Don't know for certain, though.