AI-bildgenerering😱 3-stegs guide för nybörjare🥳

Nyfiken på bildgenerering men inte testat än? Du behöver inte känna dig sen på bollen! En ny studie från The Verge visar att bara ca 20-30% använder AI för bildgenerering.

Sätt igång nu - tre nivåer, första räcker för att komma igång.

Level 1 - välj verktyg, tre alternativ:
👉 Har du ett microsoft-konto? Testa https://lnkd.in/djBvvAcw
👉 Är okej med att skapa ett nytt konto? Ett bra val är då: https://labs.openai.com/
👉 Vill du testa direkt utan att behöva logga in? Pang på: Free AI image generator: https://www.craiyon.com/

Kom igång med verktygen

I samtliga verktyg möts du av en textruta där du ska beskriva vad du vill ha genererat (prompt) och en knapp för att skapa. Enkelt att komma igång men kan vara utmanande att få precis det resultat man vill ha. Lite åt Othello-hållet: “a minute to learn, a lifetime to master”. Testa och se, du kanske är en naturbegåvning!


👇 Level up 👇


Level 2 - Hur får man det man vill ha (prompting)?
Hur ska man beskriva vad man vill ha? S.k. prompting har växt och blivit ett yrke, kanske kortlivat, men det visar att det finns utmaningar kopplat till detta. Misströsta inte, det går att komma långt med enkla riktlinjer:

1️⃣ Var specifik:
✅ "En sköldpadda som sitter på en solig fönsterbräda bredvid en kaktus i en kruka."
❌ "En sköldpadda."

2️⃣ Använd konkreta termer:
✅ "En solig strand med lugna vågor, palmer och ett randigt rött och vitt parasoll."
❌ "En lycklig plats."

3️⃣ Iterera:
✅ Om "ett slott på en kulle" inte ger det önskade resultatet kan du prova "ett stenigt slott på en grön kulle under en klarblå himmel."
❌ Upprepa gång på gång "ett slott" utan att göra ändringar, även om det inte ger det önskade resultatet.


👇 Level up 👇


#ai Level 🚀 Inspireras av dom bästa
Det finns många platser där man kan klicka på en bild och se vilken prompt som använts för att generera den. Testa att surfa till https://prompthero.com/ och antingen inspireras av bilderna som visas upp eller sök efter det du vill skapa, eller titta direkt i bings bildgenerering.

När du hittat en bild i en stil du gillar, modifiera prompten för att passa det du vill ha, klistra in i bildgenereringsverktyget. Experimentera!

Min lärdom från detta är att bra prompter är väldigt detaljerade, experter skriver en novell.

Nyfiken på att se vad du skapar - lycka till 🙌

Refining Your AI Conversation Skills

 🥳 It's easy to draw the conclusion that large language models such as ChatGPT give pretty bad output and are not very useful. But just like learning a new language, becoming proficient in AI speak "prompt engineering", requires practice and play to undertand the correlation between input (prompt) and output (answer).

I've been researching the current understanding of how to make effective prompts. Here is a summary guide to give you a flying start:

1️⃣ Be Specific - Be clear about the desired context, outcome, length, format, style, etc.
👍Write a short, inspiring poem about project management in the style of Emily Dickinson.
👎 Write a poem about project management.

2️⃣ Use Concrete Descriptions - Keep prompts concise and to the point.
👍 Write a short story about a cat's adventure in a city.
👎 Write a story about a cat.

3️⃣ Specify What to Do, Avoid What Not to Do - It's better to instruct the model on what it should do instead of what it shouldn't.
👍 The AI should recommend a TV series from IMDB without asking users for their interests or personal information.
👎 This is an AI agent that recommends TV series to a customer. Do not ask about customer interests. Do not ask for personal information.

4️⃣ Start with Instructions and Use Separators - Place instructions at the start of the prompt, use separators like ### or """.
👍Summarize the following text: """Input text here"""
👎 Input text, summarize the text.

5️⃣ Articulate Desired Output with Examples - Guide the model by providing example outputs.
👍Translate the following English phrases into French: 'Hello, how are you?' = 'Bonjour, comment ça va?' ; Now translate: 'I am learning French.'
👎 Translate 'I am learning French' into French.

6️⃣ Use an Iterative Prompting Strategy - Start by providing a task without prior examples (zero-shot)

Input: "Translate the following English sentence into French: 'I love learning new languages.'" Output: ?

If it doesn't give the desired results, try giving a few examples (one-shot or few-shot):

Example 1: Input: "Tell a joke in the following format: 'Why don't [group] [do something]?' 'Because [punchline]!'" Output: "Why don't scientists trust atoms? Because they make up everything!"

Example 2: Input: "Tell a joke in the same format." Output: "Why don't we tell secrets on a farm? Because the potatoes have eyes, the corn has ears, and the beans stalk."

Now, perform the task: Input: "Tell a joke in the same format."

👍 Experiment with various prompt engineering strategies.
👎 Give up after the first attempt.


📌 Try starting with guidelines 1-3 in mind when crafting your prompts.

📌 Give yourself a task such as getting the AI to compile a list of books to read during the summer, then experiment, experiment, and experiment until you get the results you want.

What is your experience using ChatGPT? Do you get the output you want? What do you use it for? Give examples of your best prompts?

ChatGPT in Life Science?

Having worked in the Life Science sector for many years, I was curious to explore how large language models (LLM) like ChatGPT are steadily making its way into this traditionally slow-moving sector, especially given the multitude of compliance factors that need to be considered.

LLMs are proving to be highly effective at:
📌 Understanding and processing natural language.
📌 Identifying patterns and trends in large data sets.
📌 Synthesizing complex data into meaningful, actionable insights.

These capabilities make LLMs promising tools for addressing common challenges within Life Science such as:
💊 Processing vast amounts of data, scientific literature, and patents.
💊 Accelerating drug discovery and development processes.
💊 Streamlining the design of clinical trials, creating control groups, and interpreting complex data.

It's exciting to see that major companies like AstraZeneca, Pfizer, and Sanofi are already tapping into these tools.
💊 AstraZeneca use ChatGPT to analyze scientific literature and identify new drug targets for development.
💊 Pfizer leverage ChatGPT to automate the literature review process in drug discovery and development. This not only streamlined the process but also cut down costs.
💊 Sanofi, on the other hand, is actively collaborating with AI-focused biotech firms to integrate these language models into their drug discovery process, with the goal of accelerating knowledge sharing and progress.

However, implementing these tools in such a highly regulated business sector does come many challenges, many of which are relevant for several sectors .
📌 Lack of transparency in how the LLMs operate can be a significant barrier.
📌 Training data, the question arises whether current models are trained on the right or enough data? Gigantic life science companies own an abundance of internal data that is crucial for development but is not yet included in the open models.
📌 Privacy and security, ensuring that data is not leaked, is a top concern.

How do we solve these? For transparency, initiatives such as 'explainable AI' aim to make the decision-making process of AI models more understandable. With regards to training data, collaborations between Life Science companies and AI researchers could lead to more nuanced models trained on a wider range of data. Lastly, for privacy and security, leveraging technologies like differential privacy could offer a path forward. Still some work to be done here…

🥳 We are likely to witness a surge in collaborations between AI-focused companies and traditional life science organizations. This could ultimately lead to the faster development of safer drugs, which is a win for us all!

❓ Are you working within Life Science and using large language models?
❓If you're not using it for core business functions, what do you use it for?
❓Or is it still on the strategic level?

Will AR/VR go mainstream now?

 🍏 We have seen it before, take digital music in the late 1990s for example: the market was developing but hadn't gone mainstream, then Apple entered the stage with the iPod (2001) and suddenly everyone had digital music players. More recent examples of the same pattern are the release of the iPhone, Apple Watch, and AirPods. Apple has a unique knack for when a technology is mature enough to enter the marketplace.

Apple is rumored to release its AR/VR headset this week. Will it make AR/VR mainstream?

🌍 The world is ready.
We all say that change is difficult. But we love improvements, i.e., things that do not require a behavioral change but let us do the same thing just easier and quicker.

For meetings, the difficult behavioral change from physical to digital was completed during the pandemic. Switching to VR will be simple since it's mostly an improvement. We just need to be comfortable moving from using a screen to a headset 😀

🔬 Let's zoom in very narrowly, what does AR/VR mean for project leaders?

With the generative models such as GPT-4, much of the communication problem is solved, but with a slow, text-based, user interface that is not very engaging to use. I tried using GPT-4 for running a SCRUM meeting, it did the communication part okay - but the user experience was terrible compared to a teams- or physical meeting.

🎯 Now VR could be changing that.

Imagine you, the project leader, representing yourself in the VR-space
📌 VR-you, a virtual representation powered by yourself in real time - if you smile, your avatar will smile.
📌 AI-you, a virtual representation of you powered by AI that can do some of your tasks. A clone. Say that you are traveling and can't run the daily SCRUM, then your virtual representation does it for you and updates you on the progress. Scary? Great?

It doesn't need to be a clone of you either, imagine a as-a-service company providing project leaders and charging based on the configuration. Exciting times 🚀

❓Where do you think it will end? Will AI-you run full projects? Look up the AI-CEO for the company NetDragon for a glimpse of a possible future.

❓ What will you spend your time on when AI is not only helping you with budgets and project planning, but also takes the bulk of your meetings?

🥮 Well...there is an open spot for being human in the physical space! All you leaders, start working on your baking skills and invite your team for a cozy IRL meeting.

🔍 Explore AI without Spreading Your Corporate Secrets

😱 Ever felt the icy chill of dread realizing your confidential data might be at risk while experimenting with AI? Recently, we read about Apple banning their employees from using ChatGPT due to concerns that confidential information could be leaked. It makes perfect sense and Apple is not alone.

💡 Yet, we know that generative AI models are going to play an important role for us, and the best way to stay ahead of the curve is to experiment with it. NOW.

⌛ Until the big IT providers, Google and Microsoft, roll out AI in all enterprise IT environments, there are a few ways to run pilot AI projects. Yes, we heard the announcement that Microsoft is releasing it natively in Windows 11 during the summer, but enterprises are seldom on the latest Windows release.

🚀 There are several solutions 😄

Feel save while exploring letting AI loose on your confidential documents…

🛡️ Stay super safe by running GPT locally, i.e., on a computer that is not connected to the internet. What? This sounds super nerdy, but it is not. With very limited computer skills, I managed to install privateGPT on my computer and ask questions on documents that I provided. How cool isn't that!!! It's extremely safe but the answers were not on par with say, ChatGPT-4. See https://lnkd.in/dGrpw2cd

🔒 Use the encrypted service CapeChat. It automatically encrypts your documents and censors any sensitive data, allowing your information to stay private and secure. This sounds like an opportunity to be safe and get state-of-the-art AI results. See https://lnkd.in/dpFVDhcY Do you like it?

💻 Microsoft also provide a method for deploying ChatGPT based models, which I haven't tried yet, see https://lnkd.in/dS4TknJV. Have you tried it? Does it do the magic for you?

If you are happy using publicly available information and not your own confidential data…

🖥️ Set up specific computers without access to company logins (i.e., no access to company data), use mobile internet access, and have clear instructions on what you are allowed to do and not do. Your company will likely have some publicly available information that can be used in AI experiments. This is not the safest method but quite easy to set-up.

❓What is your approach to leverage the recent developments within AI? Is it OK to use ChatGPT within your company? Is it not? Are you eager to start now? Or wait until Google/Microsoft rolls out the solution for you? 🥳