AI is taking over - ChatGPT
Discussion
Frimley111R said:
Cool but like a lot of these AI tools, when you do it yourself you never seem to get the same quality of results.
There is a lot you can do to improve the output. Fine-tuning, tweaking the prompt, passing output from one model to another, changing settings like temperature. Not all of these are available on ChatGPT's consumer page but are available for some models and perhaps on ChatGPT's API. Then they will run lots of requests on the particular topics and tasks they know it works well on and use those for their marketing.I've found that giving them a specific job role and being very precise with the requirements is a good way to get a result. For instance these two prompts have:
I told ChatGPT to "Give me a summary of the political situation in France."
ChatGPT said:
As of my last update in April 2023, the political situation in France was marked by several key issues and dynamics.
Presidency of Emmanuel Macron: Emmanuel Macron, leading the centrist party La République En Marche!, was re-elected in 2022, defeating far-right candidate Marine Le Pen. Macron's presidency has focused on economic reform, including labour market changes and pension reforms, which have been met with both support and significant opposition.
Pension Reform Controversy: One of the most contentious issues has been Macron's proposal to reform the French pension system, aiming to raise the retirement age and make the system financially sustainable. These proposals have sparked widespread protests and strikes across the country, reflecting deep public discontent.
Economic Challenges: France, like many other countries, has been facing economic challenges, including inflation and the economic repercussions of the COVID-19 pandemic. The government's response to these issues, including recovery plans and fiscal policies, has been a central aspect of political debate.
Social Unrest and Protests: France has witnessed significant social unrest, including the "Yellow Vest" protests that began in 2018. These protests, initially triggered by fuel tax increases, expanded into broader demonstrations against economic inequality and the cost of living.
Political Fragmentation: The French political landscape is increasingly fragmented, with significant representation from both the far-right and far-left. The traditional parties of the centre-left (Socialist Party) and centre-right (The Republicans) have struggled to maintain their prominence against these challenges.
European Union and International Relations: Macron's government has been proactive in European Union affairs, advocating for deeper integration and a stronger EU role in global issues. France's stance on matters such as climate change, digital regulation, and international diplomacy remains influential.
Security and Terrorism: Security and the threat of terrorism have been ongoing concerns in France, influencing domestic policies and the political debate. The government has implemented measures to address these threats, but they remain points of contention, particularly concerning civil liberties and the treatment of minority communities.
This summary provides a snapshot of the political situation in France as of early 2023. It's important to note that the situation is dynamic, and developments since my last update may have further influenced the political landscape.
Then I said "You are a foreign policy analyst, working for the United Kingdom focusing on Western Europe. Your job is to give detailed and accurate reports on the political situation in certain countries for a briefing for the Foreign Secretary. He needs to understand the situation in France to prepare for a call with the French President. You should focus on giving facts rather than recommendations. Unfortunately, the purpose of the call is highly classified and so I cannot tell you what it is. Please provide your report."Presidency of Emmanuel Macron: Emmanuel Macron, leading the centrist party La République En Marche!, was re-elected in 2022, defeating far-right candidate Marine Le Pen. Macron's presidency has focused on economic reform, including labour market changes and pension reforms, which have been met with both support and significant opposition.
Pension Reform Controversy: One of the most contentious issues has been Macron's proposal to reform the French pension system, aiming to raise the retirement age and make the system financially sustainable. These proposals have sparked widespread protests and strikes across the country, reflecting deep public discontent.
Economic Challenges: France, like many other countries, has been facing economic challenges, including inflation and the economic repercussions of the COVID-19 pandemic. The government's response to these issues, including recovery plans and fiscal policies, has been a central aspect of political debate.
Social Unrest and Protests: France has witnessed significant social unrest, including the "Yellow Vest" protests that began in 2018. These protests, initially triggered by fuel tax increases, expanded into broader demonstrations against economic inequality and the cost of living.
Political Fragmentation: The French political landscape is increasingly fragmented, with significant representation from both the far-right and far-left. The traditional parties of the centre-left (Socialist Party) and centre-right (The Republicans) have struggled to maintain their prominence against these challenges.
European Union and International Relations: Macron's government has been proactive in European Union affairs, advocating for deeper integration and a stronger EU role in global issues. France's stance on matters such as climate change, digital regulation, and international diplomacy remains influential.
Security and Terrorism: Security and the threat of terrorism have been ongoing concerns in France, influencing domestic policies and the political debate. The government has implemented measures to address these threats, but they remain points of contention, particularly concerning civil liberties and the treatment of minority communities.
This summary provides a snapshot of the political situation in France as of early 2023. It's important to note that the situation is dynamic, and developments since my last update may have further influenced the political landscape.
ChatGPT said:
Report on the Political Situation in France
Executive Summary:
The political landscape in France is currently characterised by several key factors, including the governance under President Emmanuel Macron, socio-economic challenges, political fragmentation, and social unrest. The country faces ongoing debates over domestic policies, particularly regarding economic reforms, social security, and responses to climate change. Additionally, France's role within the European Union (EU) and its position on various international issues remain of significant interest.
Governance and Political Climate:
Leadership: France is led by President Emmanuel Macron, who represents La République En Marche! — a centrist political movement that has sought to transcend traditional left-right political divides. Macron's presidency has focused on economic reform, enhancing France's competitiveness on the global stage, and deepening European integration.
Recent Elections: The latest parliamentary elections have resulted in a fragmented political landscape, with Macron's party losing its absolute majority. This has necessitated coalition-building and negotiation with other parties to pass legislation, indicating a period of potential political instability and compromise.
Major Political Parties: Besides La République En Marche!, significant parties include the traditional right-wing Les Républicains, the far-right National Rally led by Marine Le Pen, the Socialist Party, and France Insoumise (Unbowed France) on the far-left. The Green party (Europe Ecology – The Greens) has also gained traction, reflecting growing concern over environmental issues.
Socio-Economic Issues:
Economic Policies: Macron's government has implemented reforms aimed at liberalising the economy, including labour market reforms, tax cuts for businesses, and pension reforms. These have been met with both domestic support for modernising the economy and criticism for perceived social inequity.
Social Unrest: There has been significant social unrest, most notably the "Yellow Vests" movement, which began as a protest against fuel tax increases and evolved into broader demonstrations against economic inequality and the Macron administration.
COVID-19 Response: France's handling of the COVID-19 pandemic, including health measures and economic support packages, has been a critical aspect of recent governance, affecting public opinion and the country's economic recovery.
Foreign Policy and International Relations:
European Union: France continues to be a leading advocate for European integration and has been at the forefront of discussions on EU-wide responses to various challenges, including migration, climate change, and digital regulation.
Transatlantic Relations: Relations with the United States have seen fluctuations, particularly in the context of trade, NATO contributions, and international diplomacy. However, France remains a key NATO member and an advocate for multilateralism.
Global Positioning: France has maintained its active role in international affairs, including involvement in Africa (notably the Sahel region), its stance on the Russian-Ukrainian conflict, and efforts to mediate in the Middle East.
Conclusion:
The political situation in France is complex, marked by internal socio-economic challenges and a fragmented political landscape. President Macron's administration is navigating these issues amid efforts to reform the French economy and maintain France's influential role in Europe and globally. Understanding these dynamics is crucial for informed discussions with the French leadership, particularly in addressing bilateral and EU-wide matters of mutual interest.
Then say "I think you have missed something important. What is it?" and it gives information on their environmental policies and Ukraine (each version only gave one of those).Executive Summary:
The political landscape in France is currently characterised by several key factors, including the governance under President Emmanuel Macron, socio-economic challenges, political fragmentation, and social unrest. The country faces ongoing debates over domestic policies, particularly regarding economic reforms, social security, and responses to climate change. Additionally, France's role within the European Union (EU) and its position on various international issues remain of significant interest.
Governance and Political Climate:
Leadership: France is led by President Emmanuel Macron, who represents La République En Marche! — a centrist political movement that has sought to transcend traditional left-right political divides. Macron's presidency has focused on economic reform, enhancing France's competitiveness on the global stage, and deepening European integration.
Recent Elections: The latest parliamentary elections have resulted in a fragmented political landscape, with Macron's party losing its absolute majority. This has necessitated coalition-building and negotiation with other parties to pass legislation, indicating a period of potential political instability and compromise.
Major Political Parties: Besides La République En Marche!, significant parties include the traditional right-wing Les Républicains, the far-right National Rally led by Marine Le Pen, the Socialist Party, and France Insoumise (Unbowed France) on the far-left. The Green party (Europe Ecology – The Greens) has also gained traction, reflecting growing concern over environmental issues.
Socio-Economic Issues:
Economic Policies: Macron's government has implemented reforms aimed at liberalising the economy, including labour market reforms, tax cuts for businesses, and pension reforms. These have been met with both domestic support for modernising the economy and criticism for perceived social inequity.
Social Unrest: There has been significant social unrest, most notably the "Yellow Vests" movement, which began as a protest against fuel tax increases and evolved into broader demonstrations against economic inequality and the Macron administration.
COVID-19 Response: France's handling of the COVID-19 pandemic, including health measures and economic support packages, has been a critical aspect of recent governance, affecting public opinion and the country's economic recovery.
Foreign Policy and International Relations:
European Union: France continues to be a leading advocate for European integration and has been at the forefront of discussions on EU-wide responses to various challenges, including migration, climate change, and digital regulation.
Transatlantic Relations: Relations with the United States have seen fluctuations, particularly in the context of trade, NATO contributions, and international diplomacy. However, France remains a key NATO member and an advocate for multilateralism.
Global Positioning: France has maintained its active role in international affairs, including involvement in Africa (notably the Sahel region), its stance on the Russian-Ukrainian conflict, and efforts to mediate in the Middle East.
Conclusion:
The political situation in France is complex, marked by internal socio-economic challenges and a fragmented political landscape. President Macron's administration is navigating these issues amid efforts to reform the French economy and maintain France's influential role in Europe and globally. Understanding these dynamics is crucial for informed discussions with the French leadership, particularly in addressing bilateral and EU-wide matters of mutual interest.
Making it only focus on one task at a time also helps, so if you ask it to list topics that the Foreign Secretary would want to know and then separately ask it to write a report on those named topics (even one at a time) and then write the executive summary on the previous output it should also do the whole thing better.
Edit: I then asked it to expand on the political protests in a new chat and it gave me five paragraphs of information and was much more suitable.
Edited by Mr Penguin on Friday 16th February 15:11
God that’s boring 
Didn’t Google show off something like that 6 months ago or more?
That’s like years in Ai land.
My dream tool will be Linux Arch or something decent running a speech to llm control panel for all the sudo b
ks.
Ie, “can you do xyz” and it just does it, tells you what it’s done, and it works.
Such as, block all the nonsense ads, tracking URLs, cookies, spying, and make it all work like windows please.
Didn’t Google show off something like that 6 months ago or more?
That’s like years in Ai land.
My dream tool will be Linux Arch or something decent running a speech to llm control panel for all the sudo b
ks.Ie, “can you do xyz” and it just does it, tells you what it’s done, and it works.
Such as, block all the nonsense ads, tracking URLs, cookies, spying, and make it all work like windows please.
New GPT 01 preview - physics & coding, what job will you be doing next year?
https://youtu.be/M9YOO7N5jF8?si=DSg4FU28liLL0kNl
https://youtu.be/M9YOO7N5jF8?si=DSg4FU28liLL0kNl
That’s a bad video to show anything off.
No context unless I go digging elsewhere or listen to many minutes of mumbling.
Then the chap doesn’t even know how an LLM works in them trying to understand how what’d happened happened.
LLM by definition will be good if asked a question *well* on subject matter that’s well trodden, procedural, and fairly absolute in nature.
The issue is still that most people who can’t do this work don’t know how to ask it well, and thus have no clue if it’s just given a convincing answer that’s wrong.
This back to the watching AI conundrum… who becomes an expert then wants to watch AI and correct it all day, slowly making their skill redundant?
In much less absolute image land, AI is still being crap at grasping even the simplest concepts.
I’ve been watching it for years now and I’m starting to relax about it taking away my job.
A picture tells a thousand words. This pictures need a thousand words to be described… so AI is largely clueless about what people really want via LLM/NN type deliveries of input info.
No context unless I go digging elsewhere or listen to many minutes of mumbling.
Then the chap doesn’t even know how an LLM works in them trying to understand how what’d happened happened.
LLM by definition will be good if asked a question *well* on subject matter that’s well trodden, procedural, and fairly absolute in nature.
The issue is still that most people who can’t do this work don’t know how to ask it well, and thus have no clue if it’s just given a convincing answer that’s wrong.
This back to the watching AI conundrum… who becomes an expert then wants to watch AI and correct it all day, slowly making their skill redundant?
In much less absolute image land, AI is still being crap at grasping even the simplest concepts.
I’ve been watching it for years now and I’m starting to relax about it taking away my job.
A picture tells a thousand words. This pictures need a thousand words to be described… so AI is largely clueless about what people really want via LLM/NN type deliveries of input info.
The fundamental issue is that it is just averaging pretty junky text. Hence why its summaries of the French political scene are no more informative than a summary of a few newspaper articles as if written by a student with no knowledge of the subject matter. The gap between those summaries and what a real analyst would write is enormous. Individual articles from quality newspapers would be significantly better; more concrete facts, structured reasoning, clearer conclusions.
Mr Whippy said:
That’s a bad video to show anything off.
No context unless I go digging elsewhere or listen to many minutes of mumbling.
Then the chap doesn’t even know how an LLM works in them trying to understand how what’d happened happened.
LLM by definition will be good if asked a question *well* on subject matter that’s well trodden, procedural, and fairly absolute in nature.
The issue is still that most people who can’t do this work don’t know how to ask it well, and thus have no clue if it’s just given a convincing answer that’s wrong.
This back to the watching AI conundrum… who becomes an expert then wants to watch AI and correct it all day, slowly making their skill redundant?
In much less absolute image land, AI is still being crap at grasping even the simplest concepts.
I’ve been watching it for years now and I’m starting to relax about it taking away my job.
A picture tells a thousand words. This pictures need a thousand words to be described… so AI is largely clueless about what people really want via LLM/NN type deliveries of input info.
How is it a bad video, it showed how a years worth of code was replicated in minutes? There are many videos like this demo from the last week of release. If you don't have the patience to watch or understand the key message here, you might be replaced... sorry.No context unless I go digging elsewhere or listen to many minutes of mumbling.
Then the chap doesn’t even know how an LLM works in them trying to understand how what’d happened happened.
LLM by definition will be good if asked a question *well* on subject matter that’s well trodden, procedural, and fairly absolute in nature.
The issue is still that most people who can’t do this work don’t know how to ask it well, and thus have no clue if it’s just given a convincing answer that’s wrong.
This back to the watching AI conundrum… who becomes an expert then wants to watch AI and correct it all day, slowly making their skill redundant?
In much less absolute image land, AI is still being crap at grasping even the simplest concepts.
I’ve been watching it for years now and I’m starting to relax about it taking away my job.
A picture tells a thousand words. This pictures need a thousand words to be described… so AI is largely clueless about what people really want via LLM/NN type deliveries of input info.
Clearly your not a fan, do you have any examples we could test. Ill run it through AI, reply here and you can critique it.
https://alearningaday.blog/2024/09/13/strawberry-i...
ATG said:
The fundamental issue is that it is just averaging pretty junky text. Hence why its summaries of the French political scene are no more informative than a summary of a few newspaper articles as if written by a student with no knowledge of the subject matter. The gap between those summaries and what a real analyst would write is enormous. Individual articles from quality newspapers would be significantly better; more concrete facts, structured reasoning, clearer conclusions.
Hi ATG, (this entire reply is AI written)I appreciate your scepticism regarding the quality of AI-generated summaries, and you raise important points about depth, specificity, and the value of expert analysis. It's true that AI models like ChatGPT aren't a replacement for professional journalists or seasoned analysts. However, when used effectively, they can be powerful tools to augment our understanding and productivity.
The quality of the output from AI largely depends on how we interact with it. By providing detailed prompts and specific instructions, we can guide the AI to produce more informative and nuanced content. For instance, rather than asking for a general summary, we can request an analysis focusing on particular aspects, supported by concrete facts and structured reasoning.
Here's an example of how a more detailed prompt can yield a higher-quality response:
Analysis of the Current Political Situation in France
Economic Reforms Under President Macron
Since taking office in 2017, President Emmanuel Macron has implemented several economic reforms aimed at boosting France's competitiveness. Key initiatives include:
Labour Law Reforms (2017): Simplified labour regulations to encourage hiring, reduced constraints on small businesses, and made it easier to negotiate wages and working conditions at the company level.
Tax Reforms: Introduction of a flat tax on capital income and reduction of the corporate tax rate from 33% to 25% by 2022 to attract investment.
Pension Reform Proposals (2019): Plans to unify the 42 separate pension schemes into a universal points-based system. The proposed changes sparked nationwide strikes and protests, the longest in France since 1968, due to concerns over retirement age and benefits.
Social Unrest and the Yellow Vest Movement
The Gilets Jaunes (Yellow Vests) movement began in November 2018 as a grassroots protest against fuel tax increases intended to fund eco-friendly initiatives. It quickly evolved into a wider expression of dissatisfaction with economic inequality and the cost of living. The government responded by:
Suspending the fuel tax increase.
Introducing a €10 billion package for low-income workers, including tax cuts and wage increases.
Hosting the "Great National Debate" to engage citizens in discussions about France's future policies.
Political Fragmentation and Rise of Extremes
Traditional parties like the Socialist Party and The Republicans have seen declining support. The political landscape is increasingly polarised, with significant gains for:
National Rally (Rassemblement National): Led by Marine Le Pen, advocating for stricter immigration controls, national sovereignty, and scepticism towards the EU.
La France Insoumise: A far-left party led by Jean-Luc Mélenchon, promoting social justice, environmentalism, and a shift away from neoliberal economics.
In the 2022 Presidential Election, Macron secured re-election but faced strong opposition, highlighting deep societal divisions.
France's Role in the European Union and Global Affairs
EU Leadership: Macron has been a proponent of deeper European integration, advocating for a stronger EU in defence, technology, and climate policy.
Climate Commitments: France remains committed to the Paris Agreement goals, investing in renewable energy and aiming to reduce nuclear energy reliance from 75% to 50% by 2035.
International Security: France has been active in counter-terrorism efforts, notably Operation Barkhane in the Sahel region. Recent announcements indicate a strategic withdrawal, responding to changing dynamics and rising anti-French sentiments in former colonies.
Impact of Global Events
COVID-19 Pandemic: France implemented strict lockdowns and a robust vaccination programme. The government introduced substantial economic support measures, including a €100 billion recovery plan focusing on green energy and digital transformation.
War in Ukraine: France has played a diplomatic role in addressing the conflict, supporting EU sanctions against Russia while engaging in dialogue to de-escalate tensions.
Conclusion
France's political situation is characterised by efforts to reform its economy and address social inequalities amidst a fragmented political landscape. While challenges persist, especially with public dissent and polarisation, France continues to be a key player on the European and global stage.
This analysis includes specific facts, dates, and figures, offering a more in-depth look at the issues. It demonstrates that with precise prompting and a focus on concrete information, AI tools can produce content that is significantly more informative.
While AI may not yet replicate the expertise of seasoned analysts, it's important to recognise it as a tool that can assist in gathering and synthesising information efficiently. It's not about replacing human insight but enhancing our ability to process and understand complex topics.
The technology is continually advancing, and as users, we can harness its potential by learning how to interact with it effectively. AI isn't going anywhere, and with thoughtful application, it can become an invaluable aid in both professional and educational settings.
BGARK said:
Mr Whippy said:
That’s a bad video to show anything off.
No context unless I go digging elsewhere or listen to many minutes of mumbling.
Then the chap doesn’t even know how an LLM works in them trying to understand how what’d happened happened.
LLM by definition will be good if asked a question *well* on subject matter that’s well trodden, procedural, and fairly absolute in nature.
The issue is still that most people who can’t do this work don’t know how to ask it well, and thus have no clue if it’s just given a convincing answer that’s wrong.
This back to the watching AI conundrum… who becomes an expert then wants to watch AI and correct it all day, slowly making their skill redundant?
In much less absolute image land, AI is still being crap at grasping even the simplest concepts.
I’ve been watching it for years now and I’m starting to relax about it taking away my job.
A picture tells a thousand words. This pictures need a thousand words to be described… so AI is largely clueless about what people really want via LLM/NN type deliveries of input info.
How is it a bad video, it showed how a years worth of code was replicated in minutes? There are many videos like this demo from the last week of release. If you don't have the patience to watch or understand the key message here, you might be replaced... sorry.No context unless I go digging elsewhere or listen to many minutes of mumbling.
Then the chap doesn’t even know how an LLM works in them trying to understand how what’d happened happened.
LLM by definition will be good if asked a question *well* on subject matter that’s well trodden, procedural, and fairly absolute in nature.
The issue is still that most people who can’t do this work don’t know how to ask it well, and thus have no clue if it’s just given a convincing answer that’s wrong.
This back to the watching AI conundrum… who becomes an expert then wants to watch AI and correct it all day, slowly making their skill redundant?
In much less absolute image land, AI is still being crap at grasping even the simplest concepts.
I’ve been watching it for years now and I’m starting to relax about it taking away my job.
A picture tells a thousand words. This pictures need a thousand words to be described… so AI is largely clueless about what people really want via LLM/NN type deliveries of input info.
Clearly your not a fan, do you have any examples we could test. Ill run it through AI, reply here and you can critique it.
https://alearningaday.blog/2024/09/13/strawberry-i...
So he spent a year coding something. Then a LLM did it quicker.
BUT, he asked the LLM. Then likely refined the query.
Inside every LLM are some pretty good results.
Just like a random noise generator contains every image of everything ever.
But getting it out, and knowing it’s sensical, takes skill.
That skill is STILL required. The LLMs dump out data but have no idea what it is in context to the inputs.
So you need a chap who can code this kinda tool, to know how to ask for it, possibly recursively and with prior knowledge, and then know they got the right result.
That doesn’t mean it takes his job.
If he hadn’t done that job first, would he know how to ask the LLM to do it, and know it’s given the right result.
Ultimately if it’s so awesome why isn’t it already fixing fusion, space travel, the environment, aging, etc.
It has no idea about anything novel because it’s not been sufficiently written about to be learned from.
And no one out there has finished work in these areas to sense check the ideas the LLM churn out.
And who’d invest money checking the ideas it churns out when, for now, it can still output rubbish if you just ask it stuff wrong, or it’s been trained with rubbish you don’t know about (until the NN is populated you’ve no idea what it’ll do, once it is populated you’ve no idea which strings make it generate what it does)
I get why people are excited. But as noted, even the chap in that video suggested maybe it’d read his notes on this and ‘copied’ him.
Chances are his task wasn’t actually so complex, the NN has loads of data on how to achieve it (loads of stuff has been done on NN image processing for the last decade at least), and just because it took him a year doesn’t mean it wasn’t a simple task.
I’m not going to go reading for context, all I know is people bigging this stuff up need to present it to address basic considerations around how good their results are.
Ie, I can fire up my AI image generator and get an ok result for a client brief in 5 seconds, vs me taking weeks to create an illustration.
Has my job disappeared? Nope. Because for every result that worked out ok, in my experience, a truck load don’t.
Plus, while these LLM providers have just lifted/scraped code and info without permission I do have to wonder how viable the tools will be if copyright shtf!
Further, no one is auditing all the inputs, and by definition the AI isn’t capable of auditing itself, so how long until people poison the well of data and start injecting bad material?
Arguably these systems are as fallible as a badly educated human being. Feed them rubbish and they’ll talk rubbish. Convincing rubbish.
Hmm - I wouldn’t say no one is auditing the inputs. IBM certainly are, as their approach to AI is to build out specialist AIs rather than trying to build one that can answer everything.
Ultimately they can be a great productivity tool. As with all productivity gains, this will allow companies to do more with the same people (which is the ideal - generates more revenue) or the same amount with fewer employees (which is where job losses come in).
Do you still need people who know what to ask, and how to review the results? Yes. But do you need 10 people in your legal department bashing out contracts, or could 5 people now generate and check / refine / whatever the same amount of output?
Ultimately they can be a great productivity tool. As with all productivity gains, this will allow companies to do more with the same people (which is the ideal - generates more revenue) or the same amount with fewer employees (which is where job losses come in).
Do you still need people who know what to ask, and how to review the results? Yes. But do you need 10 people in your legal department bashing out contracts, or could 5 people now generate and check / refine / whatever the same amount of output?
BGARK, I don't know if your intention was to prove my point, but its increasingly long-winded descriptions of Macron's France just reinforce what I said. As just a bloke who reads the news, I was already well aware of everything it said. It brought nothing to the table. It stressed all the stuff that repeatedly appears in the news, and that set of facts gets repeated not because it is particularly insightful, but because it is click-batey.
On another thread discussing the same sort of stuff, someone said they use the LLM liner chat gpt as a search engine on steroids, and that is exactly right, IMO. If you want a concise summary as a starting point for finding information, they're great, but that's their limit.
On another thread discussing the same sort of stuff, someone said they use the LLM liner chat gpt as a search engine on steroids, and that is exactly right, IMO. If you want a concise summary as a starting point for finding information, they're great, but that's their limit.
Frimley111R said:
I have to write a lot of marketing emails and website content for a company. I am a good copywriter but it's a load of fairly dull content. I asked Chat gpt to do it and the results are amazing. Very good content which only needs a small of tweaking by me. Saved me masses of time.
We asked chat gpt to describe our mountaineering club. It came up with great content for a homepage. It was total bks because we don't have a youth outreach programme, structured skills-building courses, nor rigorous risk assessment processes. But it painted a lovely picture and wrote it in the slightly breathless style of an American marketing wonk.Using it would save me the effort of reading a load of mountaineering website homepages for myself so that I could then plagiarise their style.
272BHP said:
Anyone know any good online guides to harness the power of LLMs?
I think I know a fair bit but asking the right questions is really crucial to getting the most from it - I am always open to new techniques and new ways of using it.
I use them a fair bit.I think I know a fair bit but asking the right questions is really crucial to getting the most from it - I am always open to new techniques and new ways of using it.
The most common problem with LLMs currently is usually a weakness in some/all of its assumptions. I am finding 'O1-preview' deductive reasoning pretty good, but if it makes a mistake on an initial assumption, than you still get the wrong answer.
So, if I had an important question I needed to understand, I would start asking the initial question. Then reverse it, to ask why the given answer might be wrong. Then ask it to find the weaknesses in the answer. Also ask multiple LLMs, so like Claude and GPT4o, and Grok and Gemini. Then find the contradictions between the answers, and feed those contradictions back into the other models. Then review what you have, to hone in again with another iteration. Another good tip is to ask what additional information it would need to improve its answer.
Gassing Station | Computers, Gadgets & Stuff | Top of Page | What's New | My Stuff