- AI For All
- Posts
- DALL·E 3 Has Arrived
Hello readers,
Welcome to another edition of This Week in the Future, and what a week it was! We’ve got a lot to cover like OpenAI announcing DALL·E 3, DeepMind releasing AlphaMissense, an AI startup offering celebrities security against deepfakes, plus our coverage of VOICE & AI 2023 in DC!
As always, thanks for being a subscriber! We hope you enjoy this week’s content — for a video breakdown, check out the episode on YouTube.
Let’s get into it!
DALL·E 3 Arrives
OpenAI sparked attention this week with the announced release of DALL·E 3, their advanced text-to-image model. While there is no white paper at the moment, we do know a few interesting details:
Improvements
DALL·E 3 improves on its predecessor DALL·E 2 with higher-fidelity images. Just look at this comparison between the two models given the prompt “an expressive oil painting of a basketball player dunking, depicted as an explosion of a nebula.”
 DALL·E 2 |  DALL·E 3 |
Additionally, OpenAI claims DALL·E 3 will be able to better render text in images, a detail that has always shown up as garbled in all other models. Hands will also be much better, which AI has infamously struggled with.
Availability
Users of ChatGPT are in for a treat! DALL·E 3 will be integrated directly into the ChatGPT interface and will be available for ChatGPT Plus and ChatGPT Enterprise users this October.
OpenAI is emphasizing that unlike its competitor Midjourney, users will be able to easily get high-quality images without elaborate prompt engineering, which could make DALL·E 3 an attractive option for regular folks and business people who want to quickly generate a graphic or artwork on demand.
An Olive Branch to Artists
Notably, DALL·E 3 will decline requests to generate an image in the style of a living artist. Artists can request to have their work removed from the training dataset using this opt-out form. It wouldn’t be surprising if the same is done for authors as well given the number of lawsuits coming from that direction.
AlphaMissense Unfolds
DeepMind, renowned for their ground-breaking work in artificial intelligence, has introduced an intriguing new model named AlphaMissense. This development promises to revolutionize our understanding of genetic mutations and their implications for diseases.
Key Takeaways
Missense Mutations Explained: A missense mutation is a single-letter DNA substitution, which can have significant effects on protein function and thereby health.
The Immense Catalog
DeepMind compiled a staggering catalog of 71 million missense mutations.
Their classification model, AlphaMissense, successfully categorized 89% of all these mutations as either likely pathogenic or likely benign.
To provide some perspective, human experts have only been able to confirm 0.1% of these variants. This underscores the unprecedented scale and capability of AlphaMissense.
For the Greater Good: The entire catalog has been made publicly available for research purposes, reflecting DeepMind's commitment to collaborative scientific advancement.
DeepMind's venture into genetic research is emblematic of the transformative potential of combining AI with cutting-edge science. The vast gulf between the variants categorized by AlphaMissense and those confirmed by human experts is a testament to the leaps AI can make in fields where human capacity is limited.
Securing Your Virtual Self
An AI startup named Metaphysic is offering a glimpse into what could be the future of entertainment and branding with their Metaphysic Pro offering.
Metaphysic's Vision
Metaphysic posits that a celebrity’s likeness will soon become entirely virtual, becoming an integral part of their brand. Imagine a world where a virtual avatar, say Tom Cruise, doesn't just replicate his appearance and voice but also does all the press for his latest film simultaneously. Metaphysic offers individuals and enterprises the ability to create and secure their AI assets and then use that to flag any unauthorized reproductions like deepfakes.
Our Take
Metaphysic is charting this unexplored territory as the entertainment world is largely resisting the encroachment of AI — for now. The impacts of this are interesting to consider. A single virtualized celebrity could potentially make countless endorsements in a day, massively amplifying their reach. While an actor's name holds significant value in movies and commercials, over-saturation of their virtual presence might dilute this value. The rise of virtualized avatars could also lead to a surge in content, raising questions about the quality and diversity of what's being produced.
The Landscape of Conversational AI at VOICE & AI 2023
We attended VOICE & AI 2023 in Washington, D.C. where we got to speak directly to leading companies in the conversational AI space including PolyAI, Cognigy, PlayHT, and more. Check out our coverage of the event on YouTube!
🔥 Rapid Fire
Cisco acquired Splunk to help enterprises with AI
Windows 11 Copilot was launched
Amazon is adding AI to Alexa
Google Bard was added to more Google apps
OpenAI is inviting experts to their Red Teaming Network
Intel is bringing AI everywhere with its new processors
Oracle announced an AI voice assistant for healthcare
Databricks and MIT released a new report on data and AI
HiddenLayer raised $50M to bolster enterprise AI security
New SambaNova AI chips are competing with NVIDIA
Agility Robotics is opening first humanoid robot factory
Adept open sourced Persimmon-8B
Google updated its SEO algorithm for AI content
Writer raised $100M for enterprise gen AI
Huawei will focus on AI despite US sanctions
🎙️ The AI For All Podcast
This week’s episode featured Andrew Louder, the Founder and CEO of Louder Co, who discussed the actual steps a business can take to implement AI and start seeing rapid ROI. We also discussed how startups can use AI to do more with less and how AI will transform the workplace!
📖 What We’re Reading
A point we’ve made frequently over on IoT For All is the need to focus on actual solutions and not just the technology. In this week’s handpicked insights, Sequoia Capital expresses the same belief for generative AI, and despite the AI summer of discontent, McKinsey reports that the value train will barrel ahead.
Generative AI’s Act Two — Sequoia Capital (link)
“Act 2 will solve human problems end-to-end. These applications are different in nature than the first apps out of the gate. They tend to use foundation models as a piece of a more comprehensive solution rather than the entire solution.”
The organization of the future: Enabled by gen AI, driven by people (link)
“McKinsey research shows that gen AI could enable automation of up to 70 percent of business activities, across almost all occupations, between now and 2030, adding trillions of dollars in value to the global economy.”
💻️ AI Tools and Platforms
Dante → Advanced custom AI chatbots trained on your data
SuperAnnotate → Training data platform for enterprise AI
Truewind → AI-powered bookkeeping and finance
Replicant → Contact center automation platform
Aragon → Turn selfies into professional headshots