It came to my attention this prompt stopped working during the switch over to GPT4. I have fixed it and it works as expected again. Please give it a try and let me know what you think.
Thanks,
Pat
ChatGPT does not have the ability to read from the internet. All this is doing is fabricating information based on clues in the URL. Testing with an article with current events will prove this. If AIPRM was meant to insert the content from the web, it is not doing so. The hint should ask the user to paste the content of the article manually.
It must do a really good job of guessing. I have gotten some really good useable results from brand new sales pages and current news articles. Very little editing required. It is hard to imagine how it could know so many details from just the URL. here is the research ranger output from very recent a news article, just from inserting a URL: Main Topic: Silicon Valley Bankās investment strategies and social impact reports.
Key Points:
SVB pledged to give $5 billion to companies aiming to assist with āthe transition to a sustainable, low carbon, net zero emissions economy.ā
The bank was criticized for prioritizing ESG investing and diversity and social justice in hiring over sound investment principles.
SVB celebrated the diversity hire of Jay Ersapah, who established and is the lead for the LGBTQ+ network at the bank.
Summary: SVBās social impact reports recorded transactions and the bankās pledge to support the transition to a sustainable, low carbon, net zero emissions economy. However, the bank was criticized for prioritizing ESG investing and diversity and social justice in hiring over sound investment principles. SVB celebrated the diversity hire of Jay Ersapah, who established and is the lead for the LGBTQ+ network at the bank.
Categories:
SVBās pledge to support the transition to a sustainable, low carbon, net zero emissions economy.
SVB pledged to give $5 billion to companies aiming to assist with this transition.
Criticisms of SVBās investment strategies.
SVB was criticized for prioritizing ESG investing over sound investment principles.
The bank was also criticized for focusing on diversity and social justice in hiring over strict merit.
SVBās diversity hire.
Jay Ersapah was hired as the head of Financial Risk and Model Risk Management.
She established and is the lead for the LGBTQ+ network at the bank.
It does indeed do a really good job at guessing, plus it may have had access to sites that talked about the URL back in the data it was trained on - provided that URL was around back then. But therein lies one of the first problems - even if it had an exact scrape of a URL in the training data, it would be over a year old, and if the page has changed, it will not know it.
It is really, really simple to remove all guesswork and test this for yourself. Create a page on any domain that you own that looks like [your domain]kittens/fluffy-kittens/playing-with-toys/ but make the page have nothing at all to do with kittens or toys. Have ChatGPT write a page about the history of flint mining or something. Then use that URL for any prompt that claims to look up a file. Lo and behold, it will come up with a great page all about fluffy kittens.
Not sure you mean by āback thenā. The URL was one day old at the time I sent research ranger to look at it. The SVC collapse is a new story. just try the prompt and you will see what I mean.
GPT4 uses the exact same training corpus - the selected sources of documents - as GPT3.5 did, and that corpus was selected almost 2 years ago. The absolute most recent pages and documents it got to learn from were read over a year ago. It knows of NOTHING since then other than what it gets from prompts. Thatās what I mean by āback thenā.
OpenAI have been very clear that ChatGPT is not even connected to the internet, other than via the prompt interface.
You have still failed to explain how it extracts such precise details from just a URL. A brand new URL at that. Clearly it isnāt just making things up. If you would have looked at the prompt, you would have seen it does give the optiopn of simply imputing the raw test. But thatās up to the user. It also works by imputing just a URL. Like this 3-16- 2023 story. here is the output exactly as RR spit it out. After visiting SIG SAUER Acquires General Robotics, I have identified that the main topic of the webpage is the acquisition of General Robotics by Sig Sauer. The article discusses the key points of the acquisition, including the reasons behind it, the impact it will have on both companies, and what the future holds.
Here are my notes on the key ideas of the webpage, organized by topic:
Reasons for Acquisition:
Sig Sauerās goal to expand their presence in the defense industry
General Roboticsā expertise in unmanned systems and robotics
Impact on General Robotics:
Continued operation as a wholly-owned subsidiary of Sig Sauer
Access to Sig Sauerās resources and expertise
Retention of current management and staff
Impact on Sig Sauer:
Expansion of capabilities in unmanned systems and robotics
Access to General Roboticsā technology and intellectual property
Diversification of product offerings
Future Plans:
Continued development of unmanned systems and robotics
Exploration of new markets and opportunities
Focus on innovation and technological advancement
In summary, Sig Sauerās acquisition of General Robotics is aimed at expanding its presence in the defense industry, diversifying its product offerings, and accessing new technology and expertise. The acquisition will have a positive impact on both companies, with General Robotics continuing to operate as a subsidiary of Sig Sauer and Sig Sauer gaining access to General Roboticsā unmanned systems and robotics capabilities. The future plans involve continued innovation and exploration of new markets.
Note. I did it again 10 seconds later and got slightly different output
but still very relevant to the brand new article. Unless chatGPT knew
future events back in 2021, it must be finding current data another way.
If you do think chatGPT could predict the future, then you are sure
to love my newest prompt.
I love a good challenge. I believe AI is advancing very rapidly, What was true yesterday may not be true today. All of us are learning and none of us are experts. It is very clear to me we have some who think they are or are trying to establish themselves as the go-to guy. Thatās ok. However, crapping on other peoples work isnāt really helpful. Help us all out by clearly explaining how research ranger can extract exact details and give such a good summary of a webpage about events that just happened, new products just listed, so on so fourth? If it can do it blindly without seeing the page, then it can read my mind too. Oh wait. : ) LOL
So, just to make sure I understand you correctly, you are saying that OpenAI who actually built chatGPT donāt know how their AI works when they say it has no connection to the web? Interesting take. Iāll leave you with it.
Please do not double down on being wrong. We allow people to express their opinions, even the massively misinformed ones. But we do NOT allow people to deliberately spread false information.
That it [ChatGPT] doesnāt access the web, files, databases, other than its own is a fact.
How it manages to still sound convincing when it doesnāt have that ability is an interesting point for discussion, and very much worth spending time to dig into. But it doesnāt change the facts.
You are allowed by the forum rules to argue with the points I make, of course. Present your arguments clearly and without personal attacks or ad hominem nonsense. Otherwise you will lose access to this community and the rules (and fellow members) you disrespect.
without personal attacks or ad hominem nonsense
That is nonsense. I did no such thing. just an observation. My opinion.
Am I allowed that? If I am not welcome here, fine. Threats are not
necessary and the door will not hit me when I leave,
donāt worry about that. Misinformation? LOL
Jailbreaks can cause chatgpt to do all kinds
of things outside the official narrative of
stated capabilities. That is a fact. You may not
like that. You may not want people to know that.
But, it is a fact nonetheless. But, the official
narritive must never be challenged, right.?
Those who do must be destroyed. Go ahead.
You could kick me off AIPRM all together if you
want. It might be worth losing the lousy 20
bucks a month.
Iād say that the insinuation was personal, attacking the person rather than the points, which is forbidden in this community. The accusation of ācrapping on other peopleās workā is pretty clear too. And why? All because you were shown to be wrong and couldnāt accept it?
Iām not quite sure what to make of the latest outburst. Is this a teenage tantrum hissy-fit where you are saying āOMG you might as well lock me in my room forever then!ā because you were told you needed to finish your homework first, in your asking to be kicked from the forum and banned from using AIPRM, or a serious request?
Either way, that is not acceptable behaviour in this community, nor does it show the ability to understand or follow the rules, and thus, your ability to post further has been revoked until you can prove that you (a) understand the rules you agreed to in joining and (b) have the capability to follow such rules like an adult. Please contact AIPRM directly if you wish to go further in closing your account.