- Everyday AI
- Posts
- Are Adobe and OpenAI teaming up for Sora? š²
Are Adobe and OpenAI teaming up for Sora? š²
Apple making AI local to iPhone, UK to regulate AI models, how to build GPTs inside ChatGPT, and more!
Everyday AI
April 15, 2024
š Subscribe Here | š£ Hire Us To Speak | š¤ Partner with Us | š¤ Grow with GenAI
Outsmart The Future
Today in Everyday AI
7 minute read
š Daily Podcast Episode: Is Adobe in trouble for training its AI on other AI images? Will Google finally step up its AI game with these new AI features? Hereās this weekās AI News That Matters. Give it a listen.
šµļøāāļø Fresh Finds: Former US Treasury Secretary looks to rebuild TikTok with AI, how AI helps file your taxes, and AI that keeps your inbox organized. Read on for Fresh Finds.
š Byte Sized Daily AI News: Adobe makes big AI moves to partner with OpenAI and other AI giants, Apple to make AI local to iPhone, and the UK looking to regulate AI models. For that and more, read on for Byte Sized News.
š AI In 5: Canāt find the perfect GPT inside ChatGPT for your needs? Weāre showing you how to build your own GPT to save you time! See it here
š§ AI News That Matters: From OpenAIās data leaks to Google's slew of AI announcements, weāre breaking down this weekās AI news and why it matters Keep reading for that!
ā©ļø Donāt miss out: Did you miss our last newsletter? We talked about LLMs vs SLMs, Meta's new AI chip, and Apple's AI plan causing stock surge. Check it here!
AI News That Matters - April 15th, 2024 š°
Will Google be (more?) relevant now with its AI announcements?
Why did Adobe reportedly train its AI image model with.... AI images?
And is OpenAI in trouble with YouTube?
We've got those answers and more with our weekly AI News That Matters segment.
Join the conversation and ask Jordan any questions on AI here.
Also on the pod today:
ā¢ OpenAI's data leak and staffing changes š¬
ā¢ YouTube's CEO's concerns about OpenAI š„
ā¢ Controversies around Adobe's Firefly š¼
Itāll be worth your 28 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Hereās our favorite AI finds from across the web:
New AI Tool Spotlight ā Breezemail helps keep your inbox organized, Collectif gives you product insights in minutes, and Packify.ai helps with your packaging design.
Trending in AI - A former US Treasury Secretary plans to rebuild TikTok's algorithm with an AI partner in order to acquire TikTok.
AI in Society ā Looking to file your taxes? Intuit and H&R Block have AI assistants that can help.
AI in Healthcare - Although generative AI has already started to help in the Healthcare industry, not everyone is on board yet.
AI Startups ā AI startup Lawhive has raised $12M for its legaltech AI platform.
1. Adobeās Explores GenAI Partnership with OpenAI Sora and More š³
Adobe is developing a generative AI video model for its Firefly family to enhance Premiere Pro. These new tools will allow users to modify videos using text prompts and increase clip lengths. The release date for these features is set for "this year," with potential integrations with other AI models like Runway, Pika Labs, and OpenAIās Sora models.
These strategic partnerships aim to empower Premiere Pro users with the ability to effortlessly generate, manipulate, and refine video content through intuitive text prompts, similar to the intuitive functionality found in Photoshop's Generative Fill feature.
Adobe's collaboration with third-party AI companies is an early exploration, providing users with more options. The integration of third-party AI models is beneficial for both Adobe and AI companies, aiming to cater to creative professionals using Premiere Pro.
2. Apple to Make AI Features Local on Your iPhone š±
Apple's decision to shift towards on-device AI processing spells good news for users, as it signifies enhanced privacy and reliability. By reducing reliance on cloud servers, iPhone owners can expect faster response times and more secure data handling. With the potential integration of cloud-based AI features from tech giants like Google and Baidu, users can enjoy a seamless blend of on-device power and external capabilities. Many expect Apple's grand reveal at WWDC on June 10.
3. UK Government Plans to Regulate AI Models š¬š§
The UK government is in the process of drafting new legislation to regulate artificial intelligence technologies, focusing on large language models like OpenAI's ChatGPT. The proposed regulations aim to address concerns regarding biases, potential harms, and the need for safety testing in AI development. As concerns about the impact of AI on various aspects of society continue to rise, policymakers are exploring ways to enhance regulations to ensure responsible AI deployment.
4. Adobe Launches AI Assistant for Digital Documents š
Adobe has announced its new AI assistant that helps users understand digital documents, starting at $4.99 a month. This cutting-edge tool responds to voice commands on mobile, integrates with popular browsers, and offers features like generating summaries and providing citations from text. Adobe plans to expand its capabilities to support multiple documents simultaneously.
5. Cisco's AI Chief Shares Insights on AI in the Workplace š§āš¼ļø
Cisco's senior VP, Javed Khan, spills the tea on AI transforming work dynamics. With Cisco's AI Assistant and Webex Contact Center, efficiency and customer satisfaction are on a whole new level. Khan emphasizes the importance of balancing AI advancements with workforce preparedness in the era of remote work.
How to Build a GPT Inside ChatGPT
Canāt find the perfect GPT inside ChatGPT for your workflow?
Looking to create your own custom GPT to fit your needs?
Weāre showing you how to create custom Generative Pre-trained Transformers (GPTs) that will streamline your workflow and save you precious time!
Check out today's AI in 5.
Hands off my training data. ā
That seemed to kinda be a trend in the world of GenAI news this week.
And, will Google be able to right its AI wrongs after its huge AI announcements this week?
Letās dive in and find out with this weekās edition of AI News That Matters.
1 ā Google Cloud Next AI Announcements āļø
Could this be the AI announcement that brings Google back to AI relevancy?
Weāll see.
Over the past 6 months, Google's stock has been a bit stagnant with just 13% growth. Compared to its competitors Microsoft's 27% and Meta's 59% growth, Google has been falling behind.
One reason?
Its lackluster consumer AI offerings, Gemini release SNAFU, and lack of cohesive branding behind its AI products.
With Google Cloud Next, Google announced new AI products and offerings across most of its lineup.
From rolling out its Gemini LLM features across various apps inside of its Google Workspace, Google also announced Google Vids, an AI-powered tool that helps users create videos by generating storyboards and compiling rough drafts using various media elements.
Google also released General Availability of its Gemini 1.5 Pro model, Vertex AI agents builder, Vertex AI Model Garden, support for NVIDIAās Blackwell GPU system in Google Cloud, and more.
What it means:
Thereās no nice way to say it ā Google has dropped the AI football multiple times over the past year, and especially hard in the past 6 months since its disastrous Gemini release.
Sure, Google has still been nabbing enterprise partnerships and AI success stories left and right. But, its non-enterprise offerings (You know, the majority of us) have fallen flat on their face.
Best case scenario ā This Google Cloud Next AI refresh could help Google rebound and join the Generative AI party with Microsoft, OpenAI, Meta, Anthropic and others.
Worst case scenario ā These rollouts donāt stick, Google re-brands all their GenAI services (again) 12 times over the next 3 weeks, and the majority of consumers continue to not use Gemini and Gemini-related features.
2 ā OpenAI fires 2 after reported leaks š„
Leaks arenāt good, unless youāre Mario and Luigiās business manager.
According to a report from The Information, OpenAI recently dismissed two researchers, Leopold Aschenbrenner and Pavel Izmailov, for suspected information leaks.
What are they reportedly leaking?
Who knows. š¤·
The internal investigation revealed that both individuals were part of OpenAI's safety team, adding a layer of complexity to the company's internal dynamics.
What it means:
This is the first public shakeup since the whole Sam Altman drama, and itās anyoneās guess as to what this means.
One has to think, though, if the two who reportedly leaked info were on OpenAIās safety team, what was the info they were leaking?
Did they find something unsafe in a future model thatās currently in testing and leak it out as a warning sign?
Itās a possibility.
3 ā YouTube CEO fires warning shots šØ
Remember that super awkward interview from OpenAI CTO Mira Murati when asked how their AI video model Sora was trained?
OpenAI CTO Mira Murati says Sora was trained on publicly available and licensed data
ā Tsarathustra (@tsarnick)
9:10 PM ā¢ Mar 13, 2024
Well, YouTubeās CEO seemingly had a response.
YouTubeās CEO Neal Mohan fired warning shots, mentioning that if OpenAI had used content from YouTube to train its new video AI program, Sora, it would be considered a "clear violation" of YouTube's terms of service.
What it means:
On the surface, this might not seem like a lot. But, this is actually huge and shouldnāt be glazed over just because people arenāt talking about it much.
In short, it seems (especially given that interview!) that Sora was trained off of copyrighted materials.
Once the OpenAI vs. New York Times domino falls, we wouldnāt be surprised if YouTube vs. OpenAI became the next landmark AI case.
4 ā Adobe accused of training on AI data š
Welp, this oneās awkward.
Adobe has quietly been making long-term AI power moves by being a bit more intentional about how itās training its AI models.
Specifically, Adobe has previously stated that itās only using media in its own stock portfolio to train its AI models like Adobe Firefly.
A new report from Bloomberg says thatās not really the case, as Adobeās training data for Firefly apparently was made up of 5% of Midjourney images.
What it means:
Is this where the whole āpie in the faceā term comes from? Might be.
Hereās the problem.
Even if your training data is 95% from your own library and 5% from sources like Midjourney, it really destroys the whole āethicalā angle.
(Again, Midjourney and other AI image generators like Dall-E regularly spit out copyrighted works even when you donāt ask for them.)
Despite the training data controversy, Adobe maintains that images created with Firefly are safe to use without risking copyright infringement.
But now, you might have to take that with a 5% grain of salt?
5 ā Adobe paying creatives toā¦ replace them? š„“
Another kinda awkward moment here related to Adobe, but this one might be awkward for Adobeās users.
Recent reports show that Adobe is paying artists for video to help them train their models.
Adobe is investing in its generative AI platform, Firefly, by offering $120 to photographers, videographers and artists for submitting short video clips of everyday actions and emotions for AI training, as reported by Bloomberg.
What it means:
On one hand, thatās a great step in the right direction. Models are copyright-free, and creators are getting paid!
On the other hand, are creators just ultimately getting paid fractions of pennies on the dollar to kinda replace themselves? Getting $120 per video might seem nice.
But what about when enough creators submit enough video to make these models SO good that many of these very same creatorās clients donāt need them anymore?
ā
Numbers to watch
Over 50%
Over 50% of doctors think that generative AI will save them considerable time by quickly searching medical literature or summarizing patient data.
Now This ā¦
We wanna know!
Vote to see live results
How often do you read/watch/listen to Everyday AI? |
Reply