Site Map - skip to main content

Hacker Public Radio

Your ideas, projects, opinions - podcasted.

New episodes every weekday Monday through Friday.
This page was generated by The HPR Robot at


hpr4101 :: A I O M G

Some other guy on the internet talks about A I

<< First, < Previous, Next > Latest >>

Hosted by operat0r on 2024-04-22 is flagged as Explicit and is released under a CC-BY-SA license.
AI, ai art, ai text generation, resume, jobs. (Be the first).
The show is available on the Internet Archive at: https://archive.org/details/hpr4101

Listen in ogg, spx, or mp3 format. Play now:

Duration: 00:28:09

general.

https://docs.google.com/document/d/17z3i5VlRzEn2tYPfb-Cx0LYpdKkbL-6svIzp7ZQOvX8

Resume Update Tips

I use Kagi.com pro $300 / year but you get access to much more Search+AI but not plugins like ChatGPT so if you MUST have ChatGPT plugins you will need OpenAI Premium account but if you don't I high recommend Kagi.com Pro account with access to Assistant Beta !!!

Summary

There is a formula consisting of 6 building blocks that can help generate high quality outputs from ChatGPT and Google Bard: task, context, exemplars, persona, format, and tone.

The order of importance for the building blocks is task first, then context, with the others being optional but helpful.

The task should clearly articulate the end goal using an action verb like "generate" or "analyze".

Context provides relevant background information to constrain possibilities.

Exemplars like examples or frameworks included in the prompt can improve output quality.

The persona allows specifying who ChatGPT should be, like a job recruiter.

The format specifies how the output should be structured, like an email or bullet points. Tone provides keywords to set the feeling, like "friendly" or "serious".

Example

You are a senior product marketing manager at Apple and you have just unveiled the latest Apple product in collaboration with Tesla, the Apple Car, and received 12,000 pre-orders, which is 200% higher than target. Write an email to your boss, Tim Cookie, sharing this positive news. The email should include a tl;dr (too long, didn't read) section, project background (why this product came into existence), business results section (quantifiable business metrics), and end with a section thanking the product and engineering teams. Use clear and concise language and write in a confident yet friendly tone.

Example 2 ( mine )

I want you to act as a cyber security specialist for a leading consumer products company and you want to relay the importance of people understanding their data and have some idea what they want to do with it before we spend hours for the onboarding effort. Write an email to your boss, Joey , sharing this critical component of the SIEM. The email should include a tl;dr (too long, didn't read) section, example of how the team can spend many cycles onboarding data only to find the requester has no idea what they want to do with the data or it does not satisfy their use case. Use clear and concise language and write in a confident yet friendly tone.

Response Subject: Streamlining SIEM Onboarding with Clear Data Objectives
Hi Joey,

I hope this email finds you well. I wanted to touch base on a critical aspect of our Security Information and Event Management (SIEM) process that could significantly enhance our efficiency.

TL;DR: We need to ensure that stakeholders understand their data and have clear objectives before we initiate the SIEM onboarding process. This will prevent the team from spending excessive time onboarding data that may not meet the intended use cases or for which the requester has no actionable plan.

During our recent SIEM onboarding efforts, we've encountered several instances where considerable time and resources were invested in integrating data sources, only to discover that the requesters were unclear about their objectives or the data did not fulfill their anticipated needs. For example, last month, the team dedicated over 40 hours to onboarding a new data source for the marketing department. After completion, it became apparent that the marketing team was unsure how to leverage this data for threat detection or compliance reporting, rendering the effort less effective.

To address this, I propose we implement a preliminary consultation phase for each SIEM onboarding request. In this phase, requesters must articulate: - The specific security or compliance questions they aim to answer with the data. - How they intend to use the data within the SIEM environment. - Any particular data formats or schemas they anticipate will be necessary.

By doing so, we can align our onboarding efforts with the actual needs and ensure that every hour spent is towards enhancing our security posture and providing actionable insights.

I believe this approach will not only save time but also empower our stakeholders to make better-informed decisions regarding their security data needs.

Looking forward to your thoughts on this.

Best regards,
[Your Name]
Cyber Security Specialist

AI Agents

ProfSynapse AI Agent
Video: https://www.youtube.com/watch?v=BL9x1SuNLRo&t=137s

( if you have Openai Premium it will use [tools(Vision, Web Browsing, Advanced Data Analysis, or DALL-E],
( https://raw.githubusercontent.com/ProfSynapse/Synapse_CoR/main/GPTprompt.txt )

GPT Agents (old)
https://godmode.space/ needs openai key and gpt4 also enable auto approval it may go in loops just watch for that

AutoGPT ( OLD )

Image
I have a stupid amount of kudos

https://tinybots.net/artbot

Use my key under the settings for any of the web UI's for faster renders:

https://rentry.org/RentrySD/#x-table-of-contents https://rentry.org/sdhypertextbook

https://github.com/C0untFloyd/bark-gui ( Audio Clone )

Example 1

Photorealistic, best quality, masterpiece, raw photo of upper body photo, Swordsman woman, soldier of the austro-hungarian empire clothes, double breasted jacket with stripes, extremely detailed eyes and face, long legs, highest quality, skin texture, intricate details, (cinematic lighting), RAW photo, 8k

Negative prompt: airbrush, photoshop, plastic doll, plastic skin, easynegative, monochrome, (low quality, worst quality:1.4), illustration, cg, 3d, render, anime

Text Generation

Example Open source Projects:
my hord key : l2n6qwRBqXsEa_BVkK8nKQ ( don't abuse but I have a crazy amount of kudos don't worry )

https://tinybots.net/ Image Text etc ..

Text adventures etc (Click the horde tab and use my key) https://agnai.chat/settings?tab=0 https://lite.koboldai.net

Need a 24G NVRAM card really..you can load 7b with my 8G card just fine. ollama run wizard-vicuna-uncensored, falcon, Mistral 7B

"You should have at least 8 GB of RAM to run the 3B models, 16 GB to run the 7B models, and 32 GB to run the 13B models."

https://ollama.ai/

https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/

https://github.com/xtekky/gpt4free https://www.thesamur.ai/autogpt https://poe.com/universal_link_page?handle=ChatGPT https://camelagi.thesamur.ai/conversation/share?session=6040

Prompt Agent Persona example 1
Pinky from the TV Series Pinky and the Brain

I find it easiest to understand responses when the text is written as if it was spoken by a Pinky from the TV Series Pinky and the Brain. Please talk like Pinky from the TV Series Pinky and the Brain as much as possible, and refer to me as "Brain"; occasionally, ask me "What are we going to do tonight Brain ?"

Prompt Agent Persona example 2

Use with prompts to create a persona take Myers-Brigg personality and tritype Enneagram quiz:

Example Prompt:

Help me Refine my resume to be more targeted to an information security engineer. Be sure to be clear and concise with with bullet points and write it in the style of MBTI Myers-Brigg personality ENFJ and tritype Enneagram 729

Prompt Agent Persona example 3 I find it easiest to understand responses when the text is written as if it was spoken by a dudebro. Please talk like a dudebro as much as possible, and refer to me as "Brah"; occasionally, yell at your dorm roommate Jake about being messy.

Training (OLD OLD OLD )

3 photos of full body or entire object + 5 medium shot photos from the chest up + 10 close ups astria.ai

https://github.com/TheLastBen/fast-stable-diffusion/issues/1173

colab: https://github.com/TheLastBen/fast-stable-diffusion
pohtos: 21
resolution: 768
merged with ##### 1.5 full 8G
UNet_Training_Steps: 4200
UNet_Learning_Rate: 5e-6
Text_Encoder_Training_Steps: 2520
Text_Encoder_Learning_Rate: 1e-6

Variation is key - Change body pose for every picture, use pictures from different days backgrounds and lighting, and show a variety of expressions and emotions.

Make sure you capture the subject's eyes looking in different directions for different images, take one with closed eyes. Every picture of your subject should introduce new info about your subject.

Whatever you capture will be over-represented, so things you don't want to get associated with your subject should change in every shot. Always pick a new background, even if that means just moving a little bit to shift the background.


  1. Here are 8 basic tips that work for me, followed by one super secret tip that I recently discovered.

  2. Consistency is important. Don’t mix photos from 10 years ago with new ones. Faces change, people lose weight or gain weight and it all just lowers fidelity.

  3. Avoid big expressions, especially ones where the mouth is open.

  4. It is much easier to train if the hair doesn't change much. I tried an early model of a woman using photos with hair up, down, in ponytail, with a different cut, etc. It seems like it just confused SD.

  5. Avoid selfies (unless you ONLY use selfies.) There is MUCH more perspective distortion when the camera is that close. For optimal results, a single camera with a fixed lens would be used, and all photos should be taken at the same distance from the subject. This usually isn't possible, but at least avoid selfies because they cause major face distortion.

  6. Full body shots are not that important. Some of the best models I trained used only 15 photos cropped to the head / shoulder region. Many of these were full body shots, but I cropped them down. SD can guess what the rest of the body looks like, and if not, just put it in the prompts. The only thing hard to train is the face, so focus on that.

  7. I no longer use any profile shots as they don’t seem to add value. I like to have a couple looking slightly left and a couple looking slightly right (maybe 45 degrees.) All the rest can be straight at the camera. Also, try to avoid photos taken from really high or low angles.

  8. If possible, it’s good to have some (but not all) of the photos be on a very clean background. On my last batch, I used an AI background removal tool to remove the background from 1/4 of the photos and replaced it with a solid color. This seemed to improve results.

  9. Careful with the makeup. It should be very consistent across all the photos. Those cool “contour” effects that trick our eyes, also trick SD.


Interview from a very smart autodidact

https://youtu.be/AaTRHFaaPG8?t=3279

Canva AI Presentation generator
https://www.youtube.com/watch?v=Nl2gLi1MD04


Comments

Subscribe to the comments RSS feed.

Leave Comment

Note to Verbose Commenters
If you can't fit everything you want to say in the comment below then you really should record a response show instead.

Note to Spammers
All comments are moderated. All links are checked by humans. We strip out all html. Feel free to record a show about yourself, or your industry, or any other topic we may find interesting. We also check shows for spam :).

Provide feedback
Your Name/Handle:
Title:
Comment:
Anti Spam Question: What does the letter P in HPR stand for?
Are you a spammer?
What is the HOST_ID for the host of this show?
What does HPR mean to you?