š¤ This Week in AI
PLUS: The AI Search War Beginsāand My Rant About Hollywood and AI
Happy Friday!
The AI search war is officially on, with OpenAI, Meta, Google, and Perplexity all stepping in to offer real-time, conversational answers. This could change how we get information and maybe even how we expect it.
Iāll also dive into why Hollywood isnāt yet seeing AIās bigger opportunities.
For more frequent AI updates and insights, follow me on LinkedIn and Threads..
But first, hereās what you need to know about AI this week (clickable links appear in orange in emails and underlined in the Substack app)::
OpenAI has launched SearchGPT, its AI-powered search engine, for paid subscribers on mobile and web, with access rolling out to the free tier in the coming weeks (though usage will be capped). I got early access this weekend, and itās pretty slick š.
Their partnerships with news and data providers allows SearchGPT to add up-to-date information and new visual designs for categories like weather, stocks, sports, news, and maps.
It provides accurate, formatted answers for various queries and also cites its sources. Just tap āCitationsā and you get a full list of clickable links that take you straight to the original sources.
For many user queries, ChatGPT will now automatically search the web, but you can also manually ask it search by clicking a small web-search icon within its dialog window.
Hereās what it looks like:
The search engine wonāt pull from or cite websites or platforms that block its web crawlers, so you wonāt see citations from The New York Times, for example (who is suing OpenAI). While I expect partner publications to get better placement, as this was reportedly part of OpenAIās pitch during negotiations, the company hasnāt confirmed the exact criteria it will use to rank or prioritize sources.
They plan to improve search particularly in the areas of shopping and travel.
š OpenAI has also created a ChatGPT Chrome Extension in case you want to switch your default search engine to ChatGPT.
AND the wait is finally over (thank god!)
š You can now search through your ChatGPT web conversation historyājust hit the search icon in the top left corner. All chats (and the content inside, not just titles) are searchable.
Pro tip: rename your chats properly anyway so can find them more easily by keyword because ChatGPT doesnāt always nail it.
Now, how about those folders so I can organize my 150+ active chats?
Meta just made its first major AI news deal with Reuters. Through a multi-year partnership, Meta AIās chatbot will now give users real-time news responses across Facebook, Instagram, and WhatsApp, and Messenger citing Reuters' stories and linking to their coverage.
Reuters will be compensated, though it's unclear whether Meta can use the content to train its LLM, Llama. š¤
Meta is also entering the race to dominate AI-powered search, alongside competitors like OpenAI, Google, Microsoft and Perplexity. The company is developing its own AI search engine to reduce its reliance on Google and Microsoftās Bing, offering users conversational answers about current events through Meta AI on WhatsApp, Instagram, and Facebook.
Adobe, the global leader in creative software is going all-in on generative AI. Despite backlash from artists, the company has no plans to cater to creators opposed to AI tools, warning that those who resist AI risk falling behind. Adobeās AI tools like Firefly have become its most popular, signaling a shift in how creatives are embracing the technology.
Universal Music Group (UMG) has partnered with AI company Klay Vision to develop an "ethical" AI music generator, which respects copyright rights.
Klay claims to be building a ālarge foundational model,ā which would be a major technical feat. But since theyāre developing it privately, we wonāt know what these promises mean in practice until launch. Many AI startups make big claims, but actual capabilities often vary widely, especially as top-tier companies like OpenAI and Anthropic have set high standards for foundational models.
Meanwhile, Grammy-winning producer Timbaland has joined controversial AI music startup Suno as a strategic advisor, taking an active role in creative direction and day-to-day product development. Suno plans to launch a remix contest for Timbalandās new single āLove Again,ā offering over $100,000 in prizes and streaming opportunities for winning remixes.
Sunoās AI generates music from text prompts, with output quality thatās drawn both praise and questions, especially about its training data. Suno, along with competitor Udio, is currently facing lawsuits from major music labels accusing it of training its AI on copyrighted music without permission. Ā
Consider me tuned in. šæ
AI startups are quietly advancing in Hollywood, offering tools to streamline creative tasks. One tool converts scripts into audio, allowing production execs, writers, and people with reading or visual impairments to listen to scripts on the go. Another visualization tool lets writers turn scripts into galleries of consistent characters and locations.
But hereās the thing: As AI models continue to advance and launch new features, similar use cases will soon be addressed by top tools like ChatGPT, Claude, and Gemini, or through platforms like ElevenLabs (one of my favorites), which are building meaningful relationships and an ecosystem that supports multiple audio use cases across industries.
šØ šØšØ Iām going to have to go on another one of my rants here...
What continues to baffle me is that the entertainment industry seems fixated on using AI mainly for creative work and content generation, like conceptualization, image and video production, and VFX.
YES, itās useful for cutting costs in those areas, and for giving creators the freedom to experiment and bring ambitious visions to life without the usual constraintsābut this tech can do SO MUCH more.
Top large language models like GPT-4 and Claude are essentially the ābrainsā of generative AI, with advanced capabilities to reason, apply logic and frameworks, and solve challenging business problems.
Theyāre capable of handling complex analysis, strategic brainstorming and planning, in-depth research projects, idea generation and even tackling repetitive tasks, all of which can free up teams to focus on higher-value work.
And at a time when recent industry layoffs have left smaller teams carrying a much heavier load, this is more pressing than ever.
Building smarter workflows and streamlining operations can help ease the burden and give companies a real competitive edgeāespecially as the ability to produce high quality content across formats quickly becomes a commoditized capability.
Smaller to mid-sized companiesāthink A24, Legendary, Blumhouse, Neon, Hartbeatāhave a unique opportunity here. Without the red tape that slows larger studios, they can more quickly train their teams and establish responsible AI guidelines and practices.
This agility means their teams can get the best results from tools like ChatGPT, while larger companies that restrict these tools are left vulnerable.
Employees often end up using them on personal devices anyway, without training or guidanceāleading to inconsistent results, noticeably lower-quality work, and serious data privacy risks.
Industries like finance, big pharma, healthcare, and manufacturing have already figured this out.
Yet somehow, Iām not hearing anyone in entertainment talk about using AI this way...now if only there were someone who really understands this stuff to help connect the dots. š¤
Ok. Rant button: off.
š Claude maker Anthropic has hired its first "AI welfare" researcher to explore whether advanced AI systems might one day deserve ethical treatment, including whether they could have needs or rights that we, as creators, must consider and protect.
Theyāve also launched voice dictation for its Claude mobile app on iPhone, iPad, and Android. This feature lets you dictate instructions and questions for up to 10 minutes at a time using your voice. š
A new research study found that AI models need between 200 to 600 examples of an artistās style or a personās face in their training data to recreate recognizable features.
Importantly, these examples donāt all have to be uniqueārepeated images of the same face or style also help the AI learn to imitate it more accurately.
Still, including more diverse examples might allow the AI to produce a wider range of creative or flexible outputs.
But if AI can create someoneās style or face with just a few hundred examples, where does the boundary for ownership lie? And how can we protect intellectual property and personal privacy?
The study offers a framework for developers to avoid copyright and privacy violations by limiting how often certain content appears in training data, while helping artists evaluate when AI-generated content might be infringing on their IP.
But the biggest question about AI and copyright remains unresolved: How will the law address AI models trained on massive datasets without consent?
While the findings do offer some insights and a starting point for deeper discussions around this topic, the suggested framework feels too simplistic to address the complexities of ownership and privacy as the tech advances.
Germanyās new AI travel influencer, Emma, is facing backlash from real travel influencers, who argue AI-generated content lacks the authenticity and personal experiences that define their work. Critics fear Emma signals a shift toward cheaper, less inspired content.
But the real problem here wasnāt the use of AI. It was the lack of a clear strategy and investment in creating a compelling character, and sloppy execution.
Surely, there is room for brands to create memorable and connective AI personalities to deepen relationships with their audiencesājust think of the animated characters weāve fallen in love with.
But they need heart, and a personality that makes us care about them and want to spend time with them.
And that takes vision, imagination, and thoughtful care.
š®My prediction: AI will allow talented and soulful storytellers to bring new and lovable characters to life in ways previously unimaginable.
LinkedIn has launched Hiring Assistant, its first AI agent to help recruiters automate repetitive tasks like creating job descriptions and finding candidates. The tool is currently available to a select group of large companies including AMD, Canva and Siemens, and will roll out more widely soon.
A study from Kingās College London found that avatar therapy significantly reduces the frequency of distressing voices for people with psychosis. By creating digital avatars that represent their voices, patients can confront and take control of these experiences. Following positive results, the therapy is being considered for real-world.
Scientists in Europe have developed an AI tool that decodes pig grunts and squeals to reveal emotions like stress or contentment. The goal? To help farmers improve animal welfare, and maybe even lead to labels that let consumers choose farms that prioritize happy, healthy animals.
That's all for this week.
Iāll see you next Friday. Thoughts, feedback and questions are welcome and much appreciated. Shoot me a note at avi@joinsavvyavi.com.
Stay curious,
Avi
ššš P.S. A huge thank you to my paid subscribers and those of you who share this newsletter with curious friends and coworkers. It takes me about 8+ hours each week to curate, simplify the complex, and write this newsletter. So, your support means the world to me, as it helps me make this process sustainable.