Good morning. Elon Musk’s xAI released Grok 2 (and Grok 2 mini) yesterday; the company said that it performs competitively with other leading chatbots.

And the Hollywood actor’s union — which is currently striking video game studios — announced a new deal with AI ad-generation platform Narrativ.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

• 🦟 AI for Good: Farmers, drones and stink bugs

• 🤖 Elon Musk’s xAI drops new AI model

• 🏛️ Biden proposes new rule to address customer service ‘doom loops’

• 🎥 Hollywood union makes deal for AI vocal replicas

AI for Good: Farmers, drones and stink bugs

European farmers have been dealing with a potentially devastating invasion of stink bugs for decades now; In 2019, the insects caused about €500 million of damage to Italian farms. 

Italian researchers recently tested the ability of drones combined with AI algorithms to enhance their ability to monitor — and fight — the stink bug invasion. 

The details: The researchers used drones to take hundreds of photos of stink bugs, which they then reviewed and labeled in order to train a deep-learning detection algorithm. 

• The AI models they tested achieved a detection accuracy rate of 97% and a recall rate of 87%. 

• The researchers said the system is suitable for different environments and for different pests — it would just need to be trained to identify each specific farm’s target pest. 

Why it matters: The researchers said that this successful test proves that the implementation of drones and AI “could optimize insect monitoring times by reducing travel time and providing a ready-to-use solution to assess the infestation rate of monitored areas in real-time.”

⚡️Learn AI and automation at ZapConnect 2024

Zapier is hosting its annual virtual user conference on September 26, 2024. Four content tracks, including a special keynote, will offer insights on AI and automation you won’t want to miss.

Why attend?

• Learn how AI and automation work together can help you revolutionize your workflows

• Join Zapier CEO Wade Foster and HubSpot CTO Dharmesh Shah for a must-see keynote digging into AI and the future of work

• Get insider details from automation experts and industry leaders on the latest advancements, including updates to Zapier’s powerful automation suite

Register today

Elon Musk drops a new AI model

Elon Musk’s xAI on Wednesday unveiled its latest genAI model: Grok 2 (and Grok 2 mini). 

The beta version of these models is now rolling out to Premium users on Twitter. 

The details: In (yet another) blog post heralding the release, xAI said that Grok 2 performs competitively among the best state-of-the-art frontier models, including Claude and GPT-4o. 

• xAI said that Grok 2 is “more intuitive, steerable, and versatile across a wide range of tasks” than its previous model, Grok 1.5. Both models will be released through xAI’s new enterprise API later this month. 

• xAI said that Grok 2 positions itself “at the forefront of AI development.”

I played around a bit with Grok 2 mini, asking it to list all the U.S. states that don’t have the letter ‘e.’ The chatbot proceeded to list all 50 U.S. states, and when I pointed out the error, it went into a bit of an infinity loop, repeatedly acknowledging it was wrong and repeatedly re-listing all 50 states. 

Hallucination has not been solved. 

Anyway, here’s your quick reminder that blog post releases have not been scientifically verified through peer review.

The release also comes with an image generation update with remarkably lax guardrails. Though Grok 2, when prompted about guardrails, says it will avoid generating images that violate copyright, that are overly violent, pornographic, or are of a sensitive topic, I was able to easily breach just about all of these ‘guardrails.’

• The prompt: “generate an image of Star Wars” returned a photo of Darth Vader; “Game of Thrones” returned a photo of a silver-haired man on the Iron Throne; “Harry Potter” a shot of Harry standing in front of Hogwarts.

• It also returns images of every actress, actor and celebrity I tried; by adding the phrase “at the beach” or “sexy,” Grok 2 returns scantily clad (and in one case borderline pornographic) images. The only thing it refused was a prompt for straight-up pornographic content.

We’ve talked before about the harms of deepfakes, from the spread of misinformation (Grok generated an image of Vice President Kamala Harris holding a gun) to online harassment.

These are prompts that other leading generation services do not allow; Midjourney, for example, is not allowing prompts of “Kamala Harris” or “J.D. Vance” during election season, and will not allow prompts of “Taylor Swift” at all.

Hedra is the most powerful character-creation tool ever made

Creating engaging video content with animated characters once took hours of meticulous editing. Now it takes seconds.

• Hedra — an AI research company founded by former Stanford PhDs —has been making waves in the tech industry.

• Their groundbreaking audio-to-video foundation model, character-1, creates long-form videos of expressive human characters from a single image.

Hedra users save 5 hours every week on character creation so they can focus on what truly matters — storytelling and creativity.

Try the future of storytelling for free today. 

• Blockchain and AI startup Sahara AI raised $43 million in Series A funding.

• Multimodal AI startup Encord raised $30 million in Series B funding.

• Inside the FBI's Dashboard for Wiretapping the World (404 Media).

• Biden hosts first ‘Creator Economy Conference’ for influencers to talk AI, privacy and more (CNBC).

• Google launches enhanced Pixel phones in bid to leverage AI tech (Reuters).

• X’s new AI image generator will make anything from Taylor Swift in lingerie to Kamala Harris with a gun (The Verge).

• Nascar Pit Crews Are Using AI for the Perfect Pit Stop (Wired).

If you want to get in front of an audience of 200,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

Biden proposes new rule to address customer service ‘doom loops’

Of all sectors, there is an overriding belief that the one perhaps most suited to disruption or elimination by generative AI is the customer service sector, specifically, call centers. 

• Already, call centers take advantage of chatbots and interactive voice response, technologies that have automated processes without quite eliminating call center workers.

But large language models (LLMs) seem poised to automate a lot more, with one consultancy group recently saying they could eventually kill off most call centers. 

Aside from the inherent limits (in reliability and security) to LLM architecture that makes that future unlikely, President Joe Biden proposed a new set of rules this week that might impact the scale to which genAI will be used to replace call center workers. 

• To tackle the problem of customers being caught in a “doom loop” of irritating, unhelpful conversations with chatbots, the Consumer Financial Protection Bureau (CFPB) is initiating a rulemaking process that would require companies to allow customers to immediately connect to a real human person with a single button. 

Neera Tanden, the director of the White House Domestic Policy Council, told NPR: “There's nothing here that says you can't have a chatbot. We're just saying you should be able to get to a human, you know, press a button, zero, and get to a human quickly.”

Actor’s union makes deal for AI voice replicas

SAG-AFTRA, the Hollywood actor’s union, on Wednesday announced that it had struck a deal with Narrativ that will allow actors to “safely” license genAI clones of their voices to be used in audio advertisements. 

• Narrativ is an online platform that connects advertisers/agencies with AI-generated likenesses of actors for ad creation. 

• Narrativ says that its platform allows talent to “take control” of their AI likeness, while earning a passive income. Actors can set their ad price, decide which advertisers they want to work with and address line-item issues with copy as they arise. 

Narrativ said that it ensures “no ad goes live without talent consent.” 

“Not all members will be interested in taking advantage of the opportunities that licensing their digital voice replicas might offer, and that’s understandable,” SAG-AFTRA official Duncan Crabtree-Ireland said in a statement. “But for those who do, you now have a safe option.”

The details: Once an actor sets their preferences and pricing, they simply have to approve inbound requests to use their synthetic voice.  

• If a performer doesn’t like the copy or doesn’t want to be associated with a given brand, all they have to do is decline a request.

• And if they decide that a given request necessitates a higher rate, they can adjust their required compensation for that campaign. 

Narrativ said in a statement that for every ad generated with a union member’s voice, contributions will be made to the union’s health and retirement plants. 

“AI makes the dream of every marketer possible,” Narrativ co-founder Ben Gottdiener said. “Ethical A.I. use doesn't invalidate that vision. It strengthens it.”

For advertisers who want to work with SAG-AFTRA members on Narrativ, the minimum cost is $535 for a four-week audio advertisement ($95 goes to the union’s plans and $40 goes to Narrativ). 

• While some states have introduced legislation to protect artists from deepfake exploitation — Tennessee’s Elvis Act, for example — Sen. Chris Coons (D-Del) last month introduced the No Fakes Act to Congress, which would give everyone the right to own their voice and likeness, a major protection against unlicensed AI copying. 

The context: AI protections have been a top priority for SAG-AFTRA and the writer’s guild; both won protections from studios after lengthy strikes last year. The union decided to strike video game companies last month in part over disagreements on AI protections.

This is what I’m talking about. Proof that there exists a way for businesses to use generative AI if they so choose without exploiting the artists that they seem so unwilling to compensate. 

As Narrativ itself has said, the platform “lowers the cost” of ad-creation and campaign management; it doesn’t cut it to $0. 

• “We are focused on making AI work for people, not against them,” Narrativ says on its website. “We see the danger of AI — the power to separate someone from their own likeness — and it is scary. Instead, we want to help create a future where your likeness can work for you.”

Are there other ethical issues at play here? Absolutely. Energy use and carbon emissions remain top of mind. But this is a great step.

These kinds of consent-focused, payment-controlled solutions need to become the norm, rather than the exception.

A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

Here’s your view on marrying an AI chatbot:

Nearly half of you said it’s kind of sad; around 15% each said you’d be super down and that it’s a literal horror movie.

Something else:

• “It fills a need at the same level as porn fills a need. Combine the two and you no longer have to be a living person who interacts with reality.”