This AI newsletter is all you need #12

This issue is brought to you thanks to Scale?AI:

AI Community, you’re Invited to attend TransformX on October 19th — 21st . TransformX is the biggest AI event of the year with over 20K attendees, 80 sessions and 100+ speakers. Hear from AI and ML experts from companies like Meta, Google, Covariant, DeepMind, Open AI and more. Take part in hands-on workshops, attend expert panel discussions and listen to keynote presentations. Registration is FREE and gives you access to all recordings!?

Register now .?

What happened this week in?AI

Hottest News

1. An artist won an art competition using AI! This brings a lot of controversy both to the AI and art fields. Many people find it fantastic, and many others think it is cheating (even up to the point of sending hate mail… ). Honestly, it doesn’t really surprise me when seeing Stable Diffusion and DALLE results and what some artists do with those generated images using Photoshop to further improve realism and the quality of the results. Plus, many artists have to learn how to code and most of them don’t only use DALLE. They have to plug in multiple networks, like an upsampler model and some outpainting SOTA models. That’s an art in itself! What do you think?
2. NVIDIA RTX 3080Ti Giveaway! Once again, many YouTubers, including Yannic Kilcher and myself (What’s AI) are giving away graphics cards thanks to NVIDIA in collaboration with their GTC events. Yes, this is another chance for you to win an NVIDIA RTX 3080Ti! Also, for the GTC event in collaboration with NVIDIA! Learn more about the giveaway and how to participate in my video ?:)
3. Professional AI Whisperers have launched a marketplace for DALL-E prompts Yes, this is true. You can now buy prompts that “guarantee” good results… “PromptBase is at the center of the new trade-in prompts for generating specific imagery from image generators, a kind of meta-art market. […] the platform lets “prompt engineers” sell text descriptions that reliably produce a certain art style or subject on a specific AI platform.” How does this sound to you? Honestly, I think it may be worth it in some very narrow use cases where you really want a specific style and have tried everything without success. I have no idea. What do you think? — I’d love to know.

Most interesting papers of the?week

1. MotionDiffuse: Text-Driven Human Motion Generation with Diffusion Model The first diffusion model-based text-driven motion generation framework, responding to fine-grained instructions on body parts, and arbitrary-length motion synthesis with time-varied text prompts.
2. FETA: Towards Specializing Foundation Models for Expert Task Applications FETA, a novel benchmark for evaluating and improving Foundation (V&L) Models performance on expert data tasks (understand technical documentation, via learning to match their graphical illustrations to corresponding language descriptions).
3. In conversation with Artificial Intelligence: aligning language models with human values [DeepMind] This paper proposes interesting answers to common questions involving an agent discussing with humans and how to align such agents with us (ethically, socially…): A few of the questions discussed are: What does it mean to align conversational agents with human norms or values? Which norms or values should they be aligned with? And how can this be accomplished?

Enjoy these papers and news summaries? Get a daily recap in your inbox !

The Learn AI Together Community section!

It’s finally time… Our RTX 3080Ti Giveaway in collaboration with NVIDIA!

Once again, we are giving away an NVIDIA RTX 3080Ti to one lucky member of the Learn AI Together community for the GTC event in collaboration with NVIDIA!

We are also giving the community 10 Deep Learning Institute (DLI) codes!

Learn more about the giveaway and how to participate in the most recent post shared in the ??weekly-ai-news-video channel ??

NVIDIA GTC’s second 2022 conference will be happening September 19–22.

Attend one or more talks and send me a screenshot to enter the giveaway. Check out the ai-events channel , where we shared the talks that seemed most interesting for this community.

tl;dr:

There are only two steps:

1. Be a member of the Learn AI Together Discord community .

2. Sign up to GTC for the giveaway and don’t forget to DM me (Louis Bouchard) your screenshot during the talk you decide to attend (not now, it has to be during a live talk!)!

Meme of the?week!

Meme shared by our fantastic Roberto , from the Towards AI team.

Featured Community post from the?Discord

We love it when you share research, whether it be your own or simply interesting papers, blog posts, or code you’ve seen. This week, I personally thank Inazuma#9447 for sharing a very interesting paper with the community! Special thanks to sin yee#3513 as well for sharing another very interesting paper related to NLP !

Convert drawings into images of real people. Cool stuff. https://eladrich.github.io/pixel2style2pixel/

Inazuma — 09/08/2022

AI poll of the?week!

Join the discussion on Discord

TAI Curated?section

Article of the?week

Fake News Detection using BERT Model Python : The rise of social media has increased the impact of Fake News on our culture. People frequently believe what they read or hear to be true, and this has a huge impact on the globe, both politically and financially. The author develops a BERT Classification model to detect Fake News using Python in this post, employing a step-by-step methodology, well-stated model architecture, and instructive graphics.

If you are interested in writing for Towards AI, please sign up here and we will publish your blog to our network if it meets our editorial policies and standards. https://contribute.towardsai.net/

Lauren’s Ethical Take on “In conversation with Artificial Intelligence: aligning language models with human?values”

We love to see an ethics paper! I really like the approach that authors Atoosa Kasirzadeh and Iason Gabriel at DeepMind took for this analysis. By focusing on the linguistic basis of speech production goals and connecting ethics to these situations and outcomes, it really highlights the nuance that’s needed for applying ethics to different settings. A model’s ability to generalize is laudable in many contexts, but not every context — some require extreme specificity for success. In addition to this, determining the extent and impact that the value of the speech affects the recipient. Types of utterances only add to this complexity.

Why is this linguistic focus so important? By drawing their conclusions from ideals based on agent-to-human interactions, the goal of the speech is automatically more realistic and replicable. This is an essential foundation to develop because language models are often built with the success condition of replicating human-to-human conversation, which can be unattainable or limiting based on the goal of the chatbot. Evolving our ethical applications is just as important as the ethics themselves, as nothing takes place in a vacuum — especially not applied AI ethics!

Join me next week as I dive further into the ethical implications of this paper!

Job offers

Senior Software Engineer @ Captur (Remote, +/- 2 hours UK time)

Senior ML Engineer @ Safe Security (Remote)

ML/Algorithms Engineer @ Aurora Insight (Hybrid Remote)

Senior Staff Data Scientist @ One Concern (Remote)

Research Scientist — Speech recognition @ Abridge (Remote)

Computer Vision Scientist @ Percipient AI (Santa Clara, CA)

Research Scientist — Machine Learning @ DeepMind (London, United Kingdom)

Senior Data Scientist @ EvolutionIQ (Remote)

Interested in sharing a job opportunity here? Contact [email protected] or post the opportunity in our #hiring channel on discord!

If you are preparing your next machine learning interview, don’t hesitate to check out our leading interview preparation platform, confetti !