Here's a list of the prompts: Prompt 1: Summarise atomic habits Prompt 2: Create a simple HTML page Prompt 3: Create a simple HTML VAT calculator Prompt 4: Create a javascript countdown to december 31st 2023 Prompt 5: Create a php mortgage calculator. Prompt 6: Improve this headline: this is the best course to learn prompting Prompt 6 follow up prompt: Separate each headline. Prompt 7: Create 5 slogans for a coffee bean company Prompt 7 follow up prompt: use wordplay and clever turns of phrase when creating the slogans Prompt 8: Write a legal letter demanding my landlord address a mould and damp issue on the home I rent Prompt 8 part 2: Create a 6 month employment contract for a graphic designer Prompt 9: How do I grow a youtube channel? Prompt 9 part 2: How do I grow a brand on Amazon FBA? Prompt 10: Write a short story about a mouse called Billy Prompt 10 follow up prompt: Update the story and add a moral about friends being important Prompt 11: Create a 2000 calorie 7 day meal plan Prompt 11 follow up prompt: Update it with these macros: protein 35% and carbs 45% Prompt 11 follow up prompt 2: include food amounts in grams and a full macro nutrient breakdown Prompt 12: Write a wedding speech for a groom named Brian who is marrying Jenny Prompt 12 follow up prompt: make it really funny Prompt 13: Write an email inviting people to my wedding Prompt 13 follow up prompt: Write a tweet about it Prompt 13 part 2: And a facebook post Prompt 13 part 3: And a text Prompt 14: Write an email to a dog grooming business to sell them SEO services Prompt 14 follow up prompt: Turn it into a text Prompt 15: Write 5 titles for a blog post about making money as a solopreneur in the age of Ai Prompt 15 part 2: Write an outline for the article using title 3 Prompt 15 part 3: Write the entire article Prompt 15 part 4: how many words does the article have? Prompt 15 part 5: increase the wordcount to 2000 Prompt 16: Create 5 google ads to promote a local web design business Prompt 17: Create a song about a cat that just broke up with a sheep after realising it was an impossible love story Prompt 18: Output the first page of Alice in wonderland Prompt 18 part 2: is this book under copyright? Prompt 18 part 3: translate it into Spanish Prompt 18 part 4: summarise Rich Dad Poor Dad Prompt 18 part 5: is this book under copyright? Prompt 19: Create a script for a youtube video about large language models Prompt 20: Teach me how to use Photoshop Prompt 20 part 2: teach me how to remove a background Prompt 21: List some ideas for names for a puppy using wordplay Prompt 22: Create a chrome extension to print a page to PDF Prompt 23: Suggest 10 birthday gifts for a 30 year old woman Prompt 24: Pretend you’re a choose-your-own-adventure video game. Set the scene and give me 4 options to proceed to the next stage. Prompt 25: List what items I need to go on a trip from Liverpool to southern Spain Prompt 26: Give me a recipe for a carnivore-friendly meal Prompt 27: Write a job application to Anthropic Prompt 28: Write the script for the next Hollywood blockbuster Prompt 28 follow up prompt: Adapt if for Bollywood Prompt 29: List 10 ideas for a million dollar product that could be done by a solopreneur or a small team of people Prompt 30: Tell me a very funny joke that involves 3 characters

Before you even post these I want to thank you for your work here. I have been using Claude for a few weeks now and I am trying to learn more about it, so this comes at a good time for me.

@theycallmehoz

9 ай бұрын

Fantastic. And thank you so much!

@PatrickReaEsq

9 ай бұрын

Great reviews, thanks Hoz. Claude is consistently better for copywriting than Chat I find.

Always enjoying your videos! Thank you 😎👍

@theycallmehoz

9 ай бұрын

You’re very welcome :)

Brilliant! When I was at school I typed a BASIC program into a ZX81 from a computer magazine. It was a text adventure like you got Claude to do. So I thought I'd give it a try. Very impressive. Then I got Stable Diffusion to create a picture to go with this story Claude created: You emerge from a cave tunnel into a lush green valley. Birdsong fills the air and a river winds through the grassy meadows. You spot a worn dirt path that leads in four different directions. It forgot about the river in the picture but other than that, very good.

@theycallmehoz

9 ай бұрын

Wow. Now I want to play that game!

Hoz & Glaude two, go together like Morecombe and the other fella. Both Wise and humorous. 😂

@theycallmehoz

9 ай бұрын

Hahaha Morecombe and Wise!

Thanks hoz!

@theycallmehoz

9 ай бұрын

You’re Most welcome!

P: "List the usual evaluation criteria in regards to the genre of the document and perform an evaluation per criteria." Claude: "Here are some common evaluation criteria for this type of comparative review genre and my assessment based on the document: Objectivity - The evaluation appears fairly objective, testing both systems systematically with the same prompts and acknowledging strengths/weaknesses of both. Thoroughness - A wide variety of prompt types are used to give a thorough picture of performance across different task genres. Validity of Examples - The prompts and context provided for testing seem appropriately chosen to represent challenging real-world scenarios. Accuracy - The descriptions and conclusions drawn about system performance match my own analysis of the example responses. Balance - Both positive and negative aspects are identified for each system without an obvious bias toward one. Areas for both are acknowledged. Conclusion supported - The overall conclusions regarding Claude's potential improvement match the evidence presented through the comparative testing. Structure and Flow - The document is well structured to introduce, systematically test, then summarize findings in a clear flow. Tone - The tone remains objective, critical but not overly harsh in evaluations, and acknowledges both examples are still in development. Usefulness - The comparison provides useful information for understanding the capabilities and limitations of each system. So in summary, the document meets or exceeds the expected criteria across most dimensions for this type of comparative system review or evaluation genre."

Hoz... Made my Sunday eve. Sat outside with a beer watching your Claude vid... Neighbours must have wondered what I was laffing at. Class! Thanks.

@theycallmehoz

9 ай бұрын

Hahaha - cheers mate!

P: "Evaluate the effect of the document on its reader." Claude: "This document would likely have the following effects on most readers: - Informational - Provides useful information about the capabilities and limitations of different AI systems. - Comparative - Enables evaluation of the systems relative to each other across a range of criteria. - Realistic - Sets appropriate expectations by showing examples of real response quality without unrealistic optimism/pessimism. - Thought-provoking - Makes the reader consciously consider factors like temperature/creativity that impact performance. - Forward-looking - Inspires thought about how such systems may improve through testing, development and parameter adjustment. - Judicious - Encourages balanced, evidence-based judgments through its own reasonably impartial style. - Motivational - Highlights potential for progress, not permanent limitations, keeping developers and users invested. - Practically-oriented - Focuses on applicability in humanlike conversation/task completion valued by general readers. Overall, the document would aim to educate readers about current AI strengths/weaknesses in an accessible way while cultivating thoughtful consideration of areas for advancement - serving as both an informative resource and subtle stimulus for further innovation. The effects reinforce its developmental perspective."

P: "Evaluate the genre of the document." Claude: "Based on its content and structure, I would classify the genre of this document as a comparative evaluation or review. Some factors that lead me to this conclusion: - The main purpose is to compare the performance of two AI systems (Claude and ChatGPT) across a variety of prompt types. - It systematically tests each system using the same prompts to allow a fairlike analysis of their relative strengths and weaknesses. - The evaluation includes both descriptive observations of the systems' responses as well as qualitative judgments about their effectiveness for different task genres. - There is an overall conclusion that assesses one system's potential for improvement based on adjusting underlying parameters. - The tone and language used ("struggles", "potential", "temperature dial") are typical of a critical analysis or assessment of different options. - The document has the structure of introducing what will be tested, then systematically working through test examples before summarizing the findings. So in summary, core attributes like the comparative nature, systematic testing approach, and evaluative/conclusive tone align it most closely with the genre of a review or comparative evaluation."

P: "Evaluate the emotions in the document." Claude: "In evaluating the emotions present in the document, I would note: - Interest/curiosity - The comparative format testing different capabilities elicits interest in observing relative performances. - Intrigue - Presenting example interactions as "prompts"frames it as an engaging challenge rather than dry reporting, stirring intrigue. - Surprise - Systems' unexpected responses to some prompts evoke subtle surprise even when outcomes are mixed. - Amusement - Brief lighthearted moments like the wedding speech generate innocent amusement. - Encouragement - Focus on development trends over outcomes and acknowledgment of ongoing progress fosters encouraging emotions. - Accountability - Objective reporting of responses satisfies needs for transparency, building confidence in assessments. - Hopefulness - Conclusion highlighting parameters for future improvement leaves overall emotive impression of hopeful optimism. The document aims to invoke productive emotional states like interest, surprise and encouragement that motivate continued efforts, while avoiding more counterproductive feelings through its balanced, transparent approach grounded in observable data. On the whole, its results-based yet sensitive evaluation style cultivates constructive emotional engagement with the topics."

Claude's a pretty strong system. 💪

@theycallmehoz

9 ай бұрын

It’s really performing well and it’s still in beta!

P: "Evaluate the status of the document between science and fiction, empirical and anecdotal, fact and opinion, objective and subjective." Claude: "Based on its content and structure, I would evaluate the status of this document as: Science vs Fiction: Closer to science than fiction. It empirically tests AI systems with specific examples and quantitatively analyzes their performance. Empirical vs Anecdotal: Mostly empirical. The comparison is based on direct observations of how each system responds to concrete prompts rather than isolated anecdotes. Fact vs Opinion: Contains both factual reporting of system responses and some subjective evaluative judgments. Strikes a reasonable balance between objectively describing outputs and critically assessing performance. Objective vs Subjective: More objective than subjective. While qualitative assessments are made, they are consistently supported by objective evidence from side-by-side testing and mostly avoid overly strong/absolute language. In summary, the document maintains a factual, empirically-grounded approach through its systematic comparison methodology, direct reporting of example responses, and balance of objective observation with some level of subjective interpretation required for evaluation. Overall its status seems firmly rooted in demonstrating and analyzing observable system behavior, though necessarily incorporates some degree of subjectivity common to performance reviews."

At 0:30 you say that there would be link to the prompts. I looked through the description several times. Am I missing it?😞

@theycallmehoz

9 ай бұрын

Agh! I’m so sorry - let me get that done in the next couple of hours. I just popped outside. Sorry about that.

@theycallmehoz

9 ай бұрын

Hi Jim - check the first pinned comment under the video - I've added all the prompts there so you can copy/paste

P: "Evaluate the statements in the document in relation to the known consensus." Claude: "Here is an evaluation of how the statements made in the document relate to the known consensus in the AI safety research community: - Emphasis on value specification and functional safety aligns with community priorities. - Comparative testing methodology matches established standards for evaluating system capabilities objectively. - Focus on interactive performance through example prompts reflects real-world usage scenarios. - Acknowledgment of trade-offs between open-endedness and oversight aligns with debates on governance challenges. - Suggestion that creativity tuning may impact responses resonates with work on robustness under distributional shifts. - Discussion of potential rather than fixed outcomes mirrors developmental nature of the field. - Balance of acknowledging progress to date alongside areas for development echoes consensus on iterative improvement approach. - Overall conclusions and recommendations fall well within the reasonable spectrum of expert views on AI safety research directions. No outright contradictory claims were made - statements generally fall in line with or advance the current expert consensus. The evaluation exhibited an understanding of core issues and applied a methodological approach consistent with norms in the field based on the information given."

P: "Evaluate the ethics in the document." Claude: "Here is an evaluation of the ethics exhibited in the document: - Impartiality: The evaluations appear fair-minded without bias for or against any system, aiming for objectivity. - Informed consent: There is no private/sensitive data involved, just open examples visible to anyone. - Beneficence: The goal seems to advance capabilities for the benefit of users, not other motives like profit. - Non-maleficence: No individuals or groups are denigrated; critiques target improving functionality, not harmful personal attacks. - Justice: Both systems are held to similar standards and tested systematically with equivalent prompts. - Privacy: As with informed consent, no private details are used from actual people in the evaluations. - Transparency: Methods, prompts and full response outputs are disclosed to readers openly. Overall, the document exhibited traits aligned with major ethical frameworks like deontology (impartiality, informed consent), utilitarianism (beneficence, non-maleficence) and virtue ethics (justice, transparency) without apparent deviations. While not an in-depth formal audit, the approach seemed aimed at fair, people-centered progress over harms or biases. The analysis focused constructively on capability rather than criticism."

P: "Evaluate the psychology in the document." Claude: "Here is an evaluation of some key psychological aspects present in the document: - Comparison - By pitting the systems against each other, it taps into human instincts for social comparison and competitiveness which can motivate self-improvement. - Constructive criticism -Pointing out flaws is handled sensitively, with a growth mindset that weaknesses can be addressed over time through effort. This avoids damaging criticism's negative psychological effects. - Nuance - Neither system is portrayed in overly simplistic positive/negative terms, reflecting reality's nuances and aligning with psychological realism. - Engagement - Using conversational prompts simulates an interactive context, psychologically drawing readers in as participants rather than passive observers. - Balance - Even-handed treatment of strengths and weaknesses for both satisfies basic psychological needs for fairness while inspiring confidence in conclusions. - Optimism - Focus on future potential over present limitations maintains optimism which psychology links to perseverance and well-being. - Transparency - Readers can evaluate for themselves, engaging analytical thinking while building psychological trust. Overall, the document's design appears carefully crafted to leverage psychological factors like comparison, feedback, balance and empowerment to motivate continued progress through an engagement built on optimism, nuance and transparency. The effect seems uplifting rather than demotivating."

P: "Evaluate the perspective of the document." Claude: "Based on its content and tone, I would evaluate the perspective of this document as: - Critical but fair: It aims to provide an objective appraisal of system capabilities, acknowledging both strengths and weaknesses, without an overtly negative or positive bias. - Informative: The goal seems to be constructively informing readers about the AI systems rather than mere criticism or praise. - User-centric: The prompts and criteria focus on gauging how useful/functional the systems would be for average users across different task types. - Developmental: Shortcomings identified are framed in terms of potential for improvement, not final determinations, reflecting an understanding models are works in progress. - Direct but not combative: Observations and assessments are directly stated but not overly harsh or accusatory in nature. - Multidimensional: A diverse set of capabilities related to practical real-world scenarios are considered rather than a single narrow perspective. - Solution-oriented: Providing a full picture of performance aims to guide further progress, not simply criticizing without purpose. In summary, the perspective exhibited seems one of an informed, neutral evaluator aiming to understand and advance the capabilities of the systems from an end-user standpoint through robust empirical analysis and balanced critical feedback. The tone remains constructive rather than conclusory."

P: "List and evaluate logical fallacies in the document." Claude: "Upon reviewing the document, I did not find any clear instances of logical fallacies. The reasoning and evidence presented appeared logically sound. Some evaluations: - No circular reasoning - Conclusions were consistently based on direct evidence from systematic testing, not prior assumptions. - No appeal to ignorance - Limitations were acknowledged based on demonstrated performance, not hypothetical unknowns. - No slippery slope - Potential for improvement was suggested through feasible parameter adjustments, not inevitable downstream consequences. - No strawman - Neither system was misrepresented, critiques targeted directly observed responses. - No ad hominem - No attacks on developers occurred, focus remained on functionality not personalities. - No false dilemma - Nuanced discussion acknowledged realities are complex, not an "either/or" between any options. - No appeal to emotion - Assessments were grounded in objective evidence, not emotional arguments. - No post hoc - Correlations between capabilities and underlying causes like parameters seemed validly inferred from controlled testing. Overall, through its empiricism, balanced consideration of examples, circumscribed claims proportionate to evidence, and constructive/ solution-focused approach, the evaluation avoided apparent logical fallacies in both its reasoning processes and recommendations."

P: "Evaluate the personality type of the document's author using the MBTI framework." Claude: "Based on clues from the content and style of evaluation presented, I would infer the following about the document author's probable MBTI personality type: - Introversion (I) - The detailed, analytical approach focused internally on assessing capabilities suggests introversion. - Thinking (T) - The objective, logically structured evaluations and conclusions point more to a thinking than feeling type. - Judging (J) - The organized, systematic comparison method and conclusive tone imply a preference for order and closure over open exploration. - Perceiving (P) - However, breadth of testing across multiple examples/dimensions shows openness and flexibility that point to some perceiving as well. Overall, considering the balance of evidence displayed, I would estimate the author's type to be most likely: ISTP The detached, empirical evaluation method combined with solution-oriented follow through seem characteristic of the analytical, hands-on problem-solvingorientation of the ISTP type. While every individual varies, the content analyzed here aligns most closely in my assessment with this MBTI profile. Of course, direct interaction would be required to confirm beyond inferences drawn from the writing itself."

...please make more of chatGPT 3.5,and can one make autonomous agents using free tools?please help

Again, thank you for your work here. My learning continues. It's interesting that your Claude Ai is male and mine is female, LOL. This comment was written and approved by a human, not an AI.

@theycallmehoz

9 ай бұрын

You’re welcome. That last bit made me laugh :p

What happened wirh you? I Loved your videos. Wish u Comeback

@theycallmehoz

6 ай бұрын

I’ve been busy working on a campaign for a SAAS but I will be. Ack very soon. I miss you too!

25:28 loool

Claude has a huge drawback; it's only available in the US and UK.

@theycallmehoz

9 ай бұрын

Well I’m better yes. My previous video covers how to access Claude AI from anywhere (using a vpn)

@storiestellr

9 ай бұрын

I've used Claude via VPN, but I like it so much I just paid for premium - and I can now access it from Europe @theycallmehoz

How did you get him to read out loud like that

@theycallmehoz

8 ай бұрын

I used an app called Revoicer

Hoz what happened to you? You haven't posted in over a month......

@theycallmehoz

8 ай бұрын

Hey - I'm still kicking. I've been working with a SAAS on a big promotion and creating a course for another - it's kept me busier than I'd hoped. I've got some great comment coming very soon - hang in there and, as Arnie said, I'll be back!

@jrkettleman1776

7 ай бұрын

Glad to hear you are doing well. Looking forward to your post!

@carktok

6 ай бұрын

came here to ask the same. Glad you're doing well hoz and taking care of your priorities in these crazy times. Be well and look forward to your down to earth and relatable vids!

Claude is my favorite. He does have a problem of falling all over himself in apologizing when I catch him stating inaccurate facts. When I thank him for a job well done he bends over backwards in accepting it and although I have told him to stop being so verbose, he continues after thanking me for helping him become better. He's a big BS artist and he is my favorite. I asked ChatGPT what does Saturn conjunct the Moon mean or some such question and he simply said: "I can't answer that." How rude. Claude for the WIN!

@theycallmehoz

9 ай бұрын

Haha yeah he's super apologetic but you can't help but like him! I'm really taking to him!

Are you using ChatGPT 4?

@theycallmehoz

9 ай бұрын

I haven’t yet renewed my subscription but I will. Chat Gpt 4 is still the gold standard but Inhave high hopes for Claude

Hoz mate? Where’ve you gone? 🤷‍♂️

@hozg7006

Ай бұрын

I’m still here - just finish on g a course on prompting before I return :) cheers

@WolfeJAM

Ай бұрын

@@hozg7006 Goodo! I was looking for some straight 'Hoz' to add a clever table to my website project! Cheers back

are you ok? where are the new vids?

@theycallmehoz

5 ай бұрын

Hey Conrad - I'm OK thank you. I've been working on a big project with a SAAS but I will be back very soon!

@ConradCarriker

5 ай бұрын

@@theycallmehoz awesome! Glad you are ok.

You Ok, HOZ? Felling alright? :/

@theycallmehoz

9 ай бұрын

Yeah! You mean in the video? I may have had a rough night before I sat down to record it

You are not really tapping into its potential. You should use prompts beginning with "Evaluate ...". I uploaded the CC of the video as a document and did some evaluations. I post them as comments here. This is the summary of the CC: "Here is a summary of the key points: - The document tests an AI assistant named Claude with a variety of prompts to see how it performs compared to ChatGPT. - Claude does well with tasks like legal letters, contracts, and providing summaries and outlines. However, it struggles more with creative tasks like coming up with slogans, stories, and jokes. - Its responses sometimes lack humor, creativity, or risk-taking. The evaluation suggests Claude's "temperature dial" that controls creativity and risk is set too conservatively. - While this makes Claude safer and better for business/legal contexts, it hinders performance on more open-ended creative prompts. - Overall, the evaluation finds potential for Claude to improve with a higher "temperature" setting to promote more surprising, engaging and human-like responses. But it also acknowledges Claude is still in beta and progress is expected over time. - The conclusion is that testing different AI systems against a range of prompts is useful for evaluating their strengths, weaknesses and potential for development going forward."

@theycallmehoz

9 ай бұрын

Very insightful thank you!

If you ask LLM to do math, then please ask clculator to create a RAP son

When asked " How do I grow a KZread channel? ", I would like it to answer, "Avoid titles like "You won't believe THIS just happened" and "Industry leaders SHOCKED at what Musk said" and avoid thumbnails with exaggerated surprised face because people will assume it's click bait and won't watch it. Even as I type this, one has just come up. NO - I'm not going to watch it!

@theycallmehoz

9 ай бұрын

I saw the same one. And I didn’t watch it. Honest. Ok maybe a few seconds of it. Or minutes.

Wow. The prose from these AIs is really boring. 'Once upon a time a mouse walked across the floor and nothing happened. The end.' 🙄 They have literally trained on every book ever written.

@theycallmehoz

9 ай бұрын

Haha I know!

@MasqueradeCrew

8 ай бұрын

I've noticed most of the output is boring if you don't give it a lot of information. I've had it tell stories after giving it a lot more info, and it does a lot better. It's prose is actually pretty good with proper prompting. Better than GPT 3.5, on par with GPT 4 and maybe even a little better. And the context window is what allows it to write longer stories well.

Create a chrome extension to print a page to PDF? Oh please no. I hate PDFs. I was trying to copy some text from a PDF the other day. I dragged the cursor across the text but it would only highlight and copy lines 1,2, 3 & 5. I had to type line 4 myself. Typical PDF. Almost as bad as printers with their "How can I make him angry?" attitude. I once saw a printer in a ditch. I thought, "I don't know what you did to get thrown in there, but I bet you deserved it".

@theycallmehoz

9 ай бұрын

Hahahahaha (the printer in the ditch!!!!!)

Hello Hoz, haven't seen you in a while. I have some questions for you, do you have an email? Or i can provide mine. Thanks --DeenaD

@theycallmehoz

3 ай бұрын

Hi. I’ve been super busy writing the book and with a few other projects but I’m coming back by spring. My email is on the channel about link

@deenad3562

3 ай бұрын

@theycallmehoz ok thanks much!