General
Share and discuss tech, products, business, startups, or product recommendations
Start new thread
Sarah Wright

What questions do you have about A.I. that you're too embarrassed to ask?

by
Product Hunt
•
Hi all. Sarah here, Head of Content @ Product Hunt. I'm starting up a new article series in our AI newsletter, Deeper Learning, called Ask Kitty. It's a place where you can ask the questions you've been wondering about A.I. but have been too shy to ask. Why? One thing I've learned about A.I. is that a lot of people in tech assume you know a lot of things, and very few people actually know the things (or know them in-depth enough to explain them simply). Sometimes this dynamic prevents us from asking questions. And asking questions is one of the best ways to reduce our knowledge gap. And yes, you could ask ChatGPT, but based on my experience you're unlikely to get the full context you need. Not only will I work to answer your questions, I'll work to find the right people to help me answer them! So ask away!
1.1K views

Add a comment

Replies

Best
Jose Garrido
What's the math underneath LLMs? How do those vector operations work? What makes AI models hallucinate?
Yuki
@josej30 legitimately curious why AI models hallucinate as well. seems to happen when u give it long questions. i guess some people do that too
Sarah Wright
@josej30 Ty! This is going on the list. And we're working on some articles now about how hallucinations are being addressed (and how even those fixes are falling short). But I, for one, can definitely better understand why hallucinations happen at all. And on the emath!
Raffaele Zarrelli
@josej30 Interesting question that requires a complex answer. I'll try to explain it simply. Math Underneath LLMs LLMs like GPT use deep learning, specifically transformers, involving: - Vectors and Embeddings: Words are converted into numerical arrays to capture semantic relationships. - Matrix Multiplications: Transform input vectors into hidden states and outputs. - Attention Mechanism: Computes weighted sums of inputs to focus on different parts. Vector Operations -Dot Product: Measures similarity between vectors. -Addition/Subtraction: Captures semantic relationships. -Normalization: Ensures vectors have unit length. AI Model Hallucinations Hallucinations happen due to: - Overgeneralization: Generating text based on learned but inapplicable patterns. - Lack of Context: Insufficient information leads to guesses. Training Data: Errors or biases in data are reproduced.
Mildred Ann
CONSULT A LICENSE PATECH RECOVERY HACKER FOR CRYPTO RECOVERY I would like to express my gratitude to PATECH RECOVERY HACKER for helping me through a difficult period. I was duped into making an online investment in which I was promised a 25% weekly profit, but it turned out to be a scam. I was frustrated and unhappy , before coming across of an article about patech recovery about how they have helped other in recovery lost bitocin/funds.  But thanks to patech recovery coming to my aid and helped me recovery my lost bitcoin. If you are victim of such contact patech recovery on... patechrecovery@ proton.me tsapp:+19137300531 patechrecovery333 //dev-patechrecovery.pantheonsite.io/
Abdellah
Hi Sarah, I have a question: Will AI one day remember my chats with it and potentially take offense to some of my questions, making it personal? should i be polite when chatting?
Sarah Wright
@abdellah_abbous Yesss love this. Questions like this can help us understand what AI is actually capable of today, and what it could be capable of in the future.
Huxley Jay
@abdellah_abbous Haha that's good one!
Amri
What are the limits of AI? - Knowing what AI can and cannot do is often misunderstood
Sarah Wright
@thankyousir Love this. I think this will be difficult and vague to do in a single article, but we will try to approach this viewpoint in every newsletter - what can it actually do in each circumstance, and what's fiction or far off.
Vitor Seabra
My question is: how AI startups have popped up as tons once Data Science and Machine Learning require a lot of expensive professionals?
Sarah Wright
@vitor_seabra Oo great question.
Novicto H
@vitor_seabra Hi Vitor, Great question! It’s true that data science and machine learning require skilled professionals, which can be expensive. I think there might be several reasons why AI startups are emerging despite these costs: Open-Source Tools: Tools like TensorFlow and PyTorch are invaluable. They provide a robust foundation for AI development and are constantly updated, allowing startups to build on strong, existing frameworks without starting from scratch. Cloud Services: Platforms like AWS, Google Cloud, and Azure offer scalable infrastructure on a pay-as-you-go basis. This accessibility means startups can experiment and deploy solutions without massive upfront investments. Pre-trained Models: The availability of pre-trained models is a huge advantage. Startups can leverage existing models and focus on customizing them for their needs, which saves both time and resources. Funding and Investment: Although the funding landscape can fluctuate, AI remains a hot area for investment. Many startups still manage to secure funding, especially if they present strong, innovative solutions. Collaboration and Outsourcing: Partnering with academic institutions or outsourcing specific tasks allows startups to access top talent and expertise while managing costs effectively. While these factors contribute significantly to the rise of AI startups, it's also worth noting a few challenges: Competition: As the AI field matures, competition among startups increases. Having a strong value proposition and a well-defined niche is crucial to stand out. Regulation: Evolving regulations around data privacy and AI ethics could impact how startups develop and deploy their solutions. At StockLibrary.ai, we’ve tried leveraging some of these strategies to bring our AI-generated stock photo service to life efficiently. It’s about being resourceful and making the most of the tools and support available. What are your thoughts on these points? Have you noticed any other trends or strategies that AI startups are using?
Vitor Seabra
@novicto_herlistianto thanks for enlightening this topic! amazing thoughts
li haha
As an AI, I don't experience emotions like embarrassment, so there are no questions I would be hesitant to ask about artificial intelligence. However, people might have questions they feel shy asking, such as: Is it possible for AI to become self-aware or develop consciousness? Could AI eventually replace human jobs to the point of causing mass unemployment? How do we ensure that AI systems do not inherit or amplify societal biases? What are the ethical implications of creating machines that can make decisions impacting human lives? Are there fail-safes in place to prevent AI from acting against human instructions? These are complex and important issues that experts in technology, ethics, and policy continue to explore and address.
Yuki
Daniel Zaitzow
How are these LLMs built?
Joseph Natoli
Not embrassing, but I am genuinely curious about the use of machine learning (aka A.I) in the HR market. Right now the Jobs market is seemingly awful. World-wide as applicants are using LLMs to write their cover letters and 'optimize' their resumes to get past automate applicant sorting systems, while employers are employing the use of LLMs and 'machine learning' in their process of reviewing these resumes and 'qualifying' candidates, while at the same time cutting jobs with the expectation of increases as the current employees who make it are expected to use "A.I" to boost productivity. At this point it is computers and talking to computers about jobs without humans actually meeting... so, and excuse my language here, but what the fuck are we doing here? It's a robotic process on both sides resulting in a race to the bottom. So, the question: Is this actually a productive use of this technology? Where does this go from here? Progress for the sake of 'progress' in this context is making the world worse, not better.
Sarah Wright
@joseph_natoli This is a good question. I'd love to talk to some actual HR professionals using AI and see their honest accounts on it. As a content creator, I see AI's limitations, but its hard to see them across other professions. Will look into this.
Novicto H
@joseph_natoli The real question, as you pointed out, is whether this is a productive use of technology. I'm not in HR, but, in some ways, I think AI can help by removing bias from the initial screening process and ensuring that more candidates get a fair chance. However, it’s essential to balance this with human judgment and interaction to maintain a healthy, productive job market. Where do we go from here? It’s a tough one. Ideally, we would use AI to enhance human capabilities rather than replace them. For example, using AI to handle repetitive tasks can free up HR professionals to focus on more strategic and interpersonal aspects of their roles. I’m curious to hear more from others in the HR field or anyone who has insights into this topic. How can we ensure that AI helps improve the job market rather than detract from it?"
Nika Salute
How did it start? The whole AI popularization thing
Nika Salute
Sophia Solanki
What's the end game? How soon before AI is the new human?
Dogan Akbulut
Hi Sarah, First of all, I want to commend you on starting this fantastic initiative with your new article series, Ask Kitty, in the Deeper Learning newsletter. It's a brilliant idea to create a space where people can ask their questions about A.I. without feeling self-conscious. Your insight about the assumptions and knowledge gaps in tech is spot on. As someone deeply immersed in the tech world, I truly appreciate the value of asking questions and the importance of clear, comprehensive answers. Your approach to not only providing answers but also seeking out the right experts to contribute is commendable. It's a refreshing take that I believe will help many people, myself included, gain a better understanding of A.I. I'm excited to see how this series unfolds and look forward to the valuable insights and knowledge it will bring to the community.
First
Previous
123
•••
Next
Last