OpenAI releases DALL-E 3 with hopes to make prompt engineering obsolete

Pivot 5: 5 stories. 5 minutes a day. 5 days a week.

1. OpenAI releases DALL-E 3 with hopes to make prompt engineering obsolete

OpenAI has announced DALL-E 3, the latest iteration of its AI image synthesis model, now fully integrated with ChatGPT. The new model is designed to render images based on complex descriptions and can even handle in-image text generation, such as labels and signs. This feature addresses challenges that plagued earlier models and sets DALL-E 3 apart in the realm of AI image synthesis.

Compared to its predecessor, DALL-E 2, the new model refines small details more effectively. For instance, it can create engaging images without requiring "hacks or prompt engineering," according to OpenAI. In a head-to-head comparison with competing models like Midjourney, DALL-E 3 stands out for its ability to faithfully and convincingly render objects based on prompts, without requiring intricate adjustments.

One of the most exciting aspects of DALL-E 3 is its native integration with ChatGPT. This allows for conversational refinements to images, effectively turning the AI assistant into a brainstorming partner. The integration will be available to ChatGPT Plus and Enterprise customers starting in early October. As AI models continue to evolve, DALL-E 3's capabilities signal a new frontier in image synthesis and conversational AI, making it a development to watch closely in the AI industry.

Read the full story here 

2. Japanese researchers develop AI to translate and understand chicken vocals

A team of researchers in Japan, led by University of Tokyo professor Adrian David Cheok, claims to have developed an AI system capable of translating the clucking of chickens. The system, dubbed "Deep Emotional Analysis Learning," interprets various emotional states in chickens, such as hunger, fear, anger, and contentment.

To validate their system, the team recorded and analyzed vocal samples from 80 chickens. Collaborating with a team of eight animal psychologists and veterinary surgeons, they achieved surprisingly high accuracy in detecting the birds' emotional states. However, the researchers acknowledged that the model's accuracy may vary with different chicken breeds and environmental conditions.

While the research awaits peer review, it opens up intriguing possibilities for the application of AI in understanding animal emotions. "If we know what animals are feeling, we can design a much better world for them," Cheok said. The team hopes to adapt these AI techniques to other animals, laying the groundwork for advancements in various animal-related industries.

Read the full story here 

3. Pryon raises $100M to Supercharge AI-powered enterprise knowledge bases

Pryon, a startup developing an AI-powered platform to index and analyze enterprise knowledge bases, has secured $100 million in a funding round led by Thomas Tull’s U.S. Innovative Technology Fund. The funding aims to support Pryon's general growth, including expanding its 100-person team and scaling its strategic partnerships.

Igor Jablokov, Pryon’s founder and a former leader of IBM's multimodal AI research team, describes the platform as a "knowledge fabric." It can interface with third-party chatbots, converting data like audio, images, text, and video into searchable formats. Pryon claims to be up to 2x more accurate than Amazon's Kendra and can ingest data up to 10x faster.

Pryon has already secured a dozen large enterprise and public sector clients, including Dell, Nvidia, and Westinghouse. "Pryon is one of the few AI-native companies designed for enterprise use from its founding days," Jablokov said. The platform can meet the needs of highly regulated environments, making it a significant player in the AI industry's enterprise sector.

Read the full story here

4. GE HealthCare to develop AI-powered ultrasound imaging to combat maternal and pediatric mortality

GE HealthCare has announced plans to develop an AI-assisted ultrasound imaging tool aimed at improving healthcare access in low-and-middle-income countries. The initiative is backed by a $44 million grant from the Bill & Melinda Gates Foundation, which has a history of investing in technologies for resource-poor settings.

The technology aims to fill a critical gap in healthcare services, particularly in countries where healthcare providers may be stretched thin. Early medical intervention, facilitated by accurate imaging, can significantly reduce maternal and child mortality rates. "Proper medical imaging is key to achieving timely diagnosis, guiding treatment decisions, and improving treatment outcomes," GE HealthCare stated.

Recently, GE HealthCare, which has a $3 billion ultrasound business, has been ramping up its investments in AI. The company acquired med tech firms like Caption Health to help design this particular tool. With this new initiative, GE HealthCare demonstrates a strategic commitment to leveraging AI in addressing critical healthcare challenges, making it a significant development in the AI industry.

Read the full story here 

5. Shortwave’s AI-powered assistant lets you ask questions about your email history

Shortwave, an email client developed by former Google employees, has launched an AI-powered assistant designed to simplify email management. One of the standout features is the ability to find emails based on natural language queries. For instance, users can ask, "What was the link to the video that the client sent me about Project Cricket's bugs?" or "Summarize emails about Project Cricket from last week in bullet points."

The assistant employs a robust technology stack, including Instructor-xl and GPT 3.5 for search, Pinecone's vector database for long-term memory, MS Marco for re-ranking emails, and GPT-4 for final user-facing output. Andrew Lee, Shortwave's CEO, told TechCrunch that the assistant is designed to assist with various email-related tasks, such as composing drafts, improving writing, translating text, and even scheduling meetings by accessing the user's calendar.

With over 20,000 active users, Shortwave assures that it is not training these models on any user data. The assistant is available for all users to try, accessible via the AI icon on the top right of the Shortwave app. As AI continues to permeate everyday tasks, Shortwave's assistant represents a significant leap in automating and enhancing email management, making it a development to watch in the AI industry.

Read the full story here 

Listen to the daily Pivot 5 podcast on Spotify, Apple Podcasts, and more