It’s time to dilly, DALL-E
You might have heard of 'Portrait of Edmond de Belamy' (2018), the very first work of art created using generative adversarial networks by Paris-based arts-collective Obvious. The name Belamy is a tribute to the father of GANs, Ian Goodfellow. In French, 'bel ami' means 'good friend,' aka 'Goodfellow,' and the inspiration behind our very own newsletter –?The Belamy!??
Portrait of Edmond de Belamy, from La Famille de Belamy (2018). (Source:?Christie's Images Ltd.?)
Since then, machines are becoming more and more artistic and creative. Lately, you might have encountered bizarre images circulating on the internet like an astronaut riding a horse in space, teddy bears mixing sparkling chemicals as mad scientists, a bowl of soup that is a portal to another dimension, and whatnot??
Ironically, these images were not created by any human illustrator. Instead, it was generated by OpenAI's?DALL-E 2, a new AI system that turns text commands into realistic images and art. The outcomes are surreal and too good to be true.
A year ago, OpenAI released its text-to-image generation model based on transformer architecture –?DALL-E. The name of this model was inspired by surrealist Salvador Dali, and the robot from Walt Disney movie Wall-E.?
Compared to the previous model, DALL-E 2 generates more realistic and accurate images with 4x greater resolution. Check out everything you need to know about DALL-E?here. You can also register to try out DALL-E 2?here.?
While some of you are still on the waiting list to experiment with DALL-E 2, here are some of the alternatives you should know. Click?here?to view the list.?
Despite the advancements in how AI can generate realistic images and art, there are a few drawbacks like counting, compositionality, etc., as?highlighted?by senior tech writer?Rachel Metz?in a Twitter post.?
AIM Videos?
Stay ahead of the curve with our weekly roundup videos, featuring the latest news, developments, and advancements in AI and analytics.
Here's last week's roundup!?
Article in Focus?
If you’re on the lookout for an organization that values its people, look no further than eClerx, which ranked 3rd in “Learning & Support” in AIM’s 50 Best Firms in India for Data Scientists to Work For report.
One of India’s leading edtech startups, iNeuron, announced that it had appointed Hitesh Choudhary as the chief technology officer. With this new role, Hitesh looks to strengthen iNeuron’s technology capabilities to handle millions of leaders, alongside expanding the team.
Google AI?conducted a?study?that looks to mitigate issues with initialising using a concept jump-start?reinforcement learning?. With this,?reinforcement learning?algorithms can be bootstrapped by gradually 'rolling in' the prior policy, aka guide policy.?
Recently, Google AI introduced the Pathways language model (?PaLM?) with 540-billion parameters. Check out more details on how this new language model was trained?here?.?
Leveraging GPT-3, this AI expert from Dubai, UAE, has built an AI model that?plays chess?. He has tricked the model into treating chess as a text generation task. Find more details on how he got GPT-3 to play chess?here.
?
People & Tech
AIM caught up with Siemens Digital Industries Software's India head Ajinkya Bhave to understand some of the rising concerns in artificial intelligence, the future roadmap, and more. Check out the interview?here?.?
"I was so inexperienced that I didn't know how to insert a floppy disk into the computer," said HCL's AI and data science head Aruna Pattam -- in an exclusive interview with AIM. Check out her full interview?here?.?
Whether it is about deciding the size of the login button on a website or making multi-million dollar decisions, experimentation has become a major tool in the hands of business owners.
领英推荐
Data Science Corner?
In a short period, Rajat Ranjan, a data scientist at TheMathCompany, has built serious street cred in data science and analytics. Check out his data science journey?here?.?
Kaggle Grandmaster Mobassir Hossein currently works at Bengali.ai, a community working on open-sourced datasets for NLP and computer vision research. Check out his inspiring journey into data science?here?.?
Startups, Apps & Tools?
Based in New Delhi and Palo Alto,?DataNeuron?helps in accelerating and automating human-in-loop labelling for developing AI solutions.
DataQA is a tool to search and label documents for various NLP tasks like entity extraction, entity linking, etc.?
A web-based, AI-powered 3D animation editor and motion capture tool.?
Research & Papers?
In this paper, researchers from Meta AI and Sorbonne University revisits the supervised training of vision transformers (ViTs). Check out the paper?here?.?
In this paper, the researchers from Duke University, University of Bari Aldo Moro, and Università di Bologna analyzed the effect of a policy recommendation on the performances of an artificial interbank market. Check out the paper?here?.?
Dialogues????
Event in Focus?
Intel?, in collaboration with AIM, brings to you oneAPI AI Analytics workshop to accelerate Python for data science and machine learning, on April 22, 2022, from 3:00 PM to 5:00 PM. Click?here?to register for this event.?
Top News, Features & Stories of the Week?
Human I Poet I Short story writer I Self-published author I Artist I Amateur Prompt Engineer, Associate Consultant at Tata Consultancy Services, @TCS Interactive | Patent Drafting I Content Management
2 年This is a good experiment that's begun. We need to see how many datasets can be provided for sustained reliability. For abstract terms synthetic data may be needed. Voice to image conversion is also a good idea.