Good morning. Last week, the New Jersey Department of Labor, in collaboration with USDR and Google.org, released a set of LLM training tools designed to turn off-the-shelf generative AI models into bilingual translation assistants for the world of unemployment insurance.

I sat down with some of the folks behind the launch — you can give it a watch (or listen) here.

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

In today’s newsletter:

• 🐳 AI for Good: Identifying beluga whales

• 🚘 GM shutters self-driving unit 

• 🏛️ Google asks FTC to go after Microsoft & OpenAI 

• 💻 Google’s Gemini 2.0 era is here

AI for Good: Identifying beluga whales

The Cook Inlet beluga whale population declined by some 80% between 1979 and 2018. In an effort to save the species, scientists at NOAA in 2020 began experimenting with artificial intelligence — specifically, deep learning — as a means of more closely monitoring the population. 

The details: One of the main ways scientists have traditionally monitored whale populations involves underwater acoustic monitoring, where underwater microphones collect audio data. 

• But the algorithms they used to use to parse through the hours of recordings struggled with noisier areas, meaning the identification process required months of labor. 

• New machine learning models are enabling the scientists to process their “backlog of raw data,” identifying whale sounds at a high rate of accuracy — and far more quickly — than human scientists could ever manage. 

“The results have been beyond expectation. Machine learning is achieving more than 96% accuracy in classifying detections compared to a scientist doing the classification,” Manuel Castellote, a NOAA Fisheries who led the study, said at the time. “It is even picking up things human analysts missed. We didn’t expect it to work as well as humans. Instead, it works better.”

Tired of Battling Spam Calls on Your Phone? Here's How to Make Them Disappear.

Every day, your personal data, including your phone number, is sold to the highest bidder by data brokers. This leads to annoying robocalls from random companies and, worse, makes you vulnerable to scammers.

Meet Incogni: your solution against robocalls. It actively removes your personal data from the web, fighting data brokers and protecting your privacy. Unlike other services, Incogni targets all data brokers, including those elusive People Search Sites.

Put an end to those never-ending robocalls and email spam on your iPhone now.

Incogni protects you from identity theft, spam calls, increased health insurance rates, and more.

Only for The Deep View readers: Get a 58% discount on Incogni's annual plan using code: DEEPVIEW

GM shutters self-driving unit 

General Motors this week officially shuttered Cruise, its robotaxi unit, saying that it plans to combine Cruise with GM’s own internal technical teams to help advance GM’s advanced driver assist and driverless technology. 

The details: GM, which currently holds a 90% ownership stake in Cruise, said it would no longer fund Cruise’s robotaxi efforts due to the significant time and cost associated with the endeavor, “along with an increasingly competitive robtoaxi market.” 

• GM said in a statement that it has plans with other shareholders to boost its ownership to 97%. GM plans to acquire the remaining shares, and then, so long as the board approves, will set about restructuring the company next year. 

• Though an early entrant in the race to driverless taxis, Cruise hit a couple of big snags last year; after an incident last year in which a Cruise robotaxi struck and dragged a pedestrian 20 feet, the company halted its operations. It has spent the past year slowly making its way back into operation. 

Deepwater’s Gene Munster called the move “good news for Tesla and Google,” which owns Waymo. “This is further evidence that the robotaxi market will be winner take most,” he said. “The likely winners are Tesla and Waymo. The losers will likely be Uber and Lyft.”

According to a statement, GM remains “fully committed to autonomous driving.”

Waymo, the self-driving unit owned by Google, remains far ahead of the competition. The firm has been steadily expanding its operations for months, and plans to bring its robotaxis to the east coast (Miami, Florida) by 2026. 

Tesla — whose Autopilot and Full-Self Driving features function as advanced driver-assist programs, not self-driving programs — plans to launch a robotaxi model in 2026. 

• Apple’s latest IOS update includes the long-awaited ChatGPT integration with Siri. When a user asks a question that Apple’s software decides is better suited for ChatGPT than its internal systems, it will ask the user permission to send the question to GPT-4o for an answer.

• Google is entering into a partnership with Intersect Power and TPG Rise Climate to build new industrial parks that will include massive data centers directly alongside clean energy power plants to power them. The first project is expected to be operational by 2026.

• Microsoft’s growing AI health ambitions (Semafor).

• OnlyFans models are using AI impersonators to keep up with their DMs (Wired).

• Tesla shares climb to record, boosted by 64% pop since Trump election victory (CNBC).

• AI enters Congress: Sexually explicit deepfakes target women lawmakers (The19th).

• Big Tech led a stock surge on Wednesday, pushing the Nasdaq above 20k (Investopedia).

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.

• Staff Researcher: Github, Remote

• Engagement Manager, GenAI: ScaleAI, San Francisco, CA

Google asks FTC to go after Microsoft & OpenAI

Google, according to The Information, recently asked the U.S. Federal Trade Commission (FTC) to break up Microsoft’s agreement with OpenAI. The specific point of contention, according to the report, involves the exclusive cloud-hosting arrangement Microsoft has with OpenAI, which prevents other cloud providers (such as Google) from hosting OpenAI’s tech. 

Google declined a request for comment. 

The details: The conversation reportedly came as part of the FTC’s broader investigation into the relationship between Microsoft and OpenAI, which has involved the investment of $13 billion by Microsoft, the aforementioned exclusive cloud hosting arrangement and a 20% cut of OpenAI’s revenue. 

• Though OpenAI — through its ChatGPT interface, which recently hit 300 million weekly active users — has been the clear dominant in the generative AI market so far, recent reports indicate that it is steadily losing market share. 

• Between 2023 and 2024, OpenAI’s market share fell some 16% to 34%. Anthropic’s market share, meanwhile, doubled to 24%. 

A changing FTC: President-elect Donald Trump said Tuesday that Andrew Ferguson, a Republican FTC commissioner, would replace current FTC head Lina Khan. Khan has spent the past few years leading aggressive regulatory efforts that have encompassed the prevention of corporate acquisitions and antitrust investigations into the bulk of the Big Tech players. 

“Andrew has a proven record of standing up to Big Tech censorship, and protecting Freedom of Speech in our Great Country,” Trump wrote on Truth Social. “Andrew will be the most America First, and pro-innovation FTC Chair in our Country’s History.”

Wedbush analyst Dan Ives was thrilled that Khan will soon be out, which, he said, “means deal making in the tech world is about to significantly increase heading into 2025.” 

Google’s Gemini 2.0 era is here

Google on Wednesday heralded the beginning of its Gemini 2.0 era with the launch of Gemini 2.0 Flash, a newly accessible “workhorse” model, alongside the unveiling of a number of research projects, all of which are focused on making so-called “agents” real. 

The details: A text-based version of Gemini 2.0 Flash, Google said, is now available to Gemini and Gemini Advanced users; Google said that, this week, it started using Gemini 2.0 to handle its AI Overviews search function, and added that the model will be rolling out to other products next year. 

• CEO Sundar Pichai called it Google’s “most capable model yet …  If Gemini 1.0 was about organizing and understanding information, Gemini 2.0 is about making it much more useful,” he said in a statement. 

• The model, according to Google, which is capable of processing image, video, text and audio data, outperforms Google’s prior models on a number of benchmarks, including code generation and speech translation. 

The improvements here are relatively incremental. Pichai recently acknowledged that progress in AI is “going to get harder … the low-hanging fruit is gone. The hill is steeper.”

An experimental, multimodal version of Gemini 2.0 is available now to developers; Google said it — along with an array of different model sizes — will become widely available in January. 

Google did not disclose the data used to train the model, nor did it disclose the energy cost or carbon emissions associated with both training and deploying the new model. The size of the model, likewise, remains unclear. Google declined to comment on these points. 

A spokesperson told me instead that Google believes “AI can be a big contributor to alleviating the climate crisis with: better models for prediction and monitoring, optimizing existing infrastructure, and accelerating scientific breakthroughs.”

The company, alongside other Big Tech peers, is exploring nuclear power as an answer to the energy requirements of its AI projects.

“It’s an important challenge to tackle, and one we take very seriously, not just for AI models … we are committed to being as carbon-efficient as possible,” the spokesperson added.

Google also gave a rundown of a few of its agentic projects — Marine, Astra and Jules — all of which will leverage Gemini 2.0 to function as a better “assistant.” 

Google said the safety risks associated with decision-making, action-taking generative AI are the reason behind its “exploratory and gradual” approach to the development and deployment of these systems. The company said it will continue to prioritize “making safety and responsibility a key element of our model development process as we advance our models and agents.”

The blogs regarding Gemini 2.0 posted by Google were not peer-reviewed, and so have not been independently verified.

I would really like to know the details of the energy cost differential between normal Google Search and Google’s AI Overviews. I think this differential is an important piece of information to include for every application that gets a generative AI makeover.

It’s all well and good to talk about safety and responsibility, but it does feel strange to pair that with a lack of full transparency. 

What we’re left with is a question of trust. And that trust is colored by the profits that drive the companies who are in turn imploring us to trust them.

🤔 Your thought process:

Selected Image 1 (Left):

• “Image 1 looks more like what I've seen in documentaries with beekeepers on national geographic, and image 2 looked really wet and goopy.”

Selected Image 1 (Left):

• “Bees are better at sustaining the regular hexagonal arrangement, and don't layer cells on top of one another.”

💭 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 the latest Character lawsuit:

26% of you said there’s no way to keep a chatbot like this safe; you’re keeping your kids far away from stuff like this. 23% said the app should be shut down; 21% said it needs age restrictions and 17% said it seems like there are a lot of other factors going on here.

Shut it down:

• “Any AI that's programmed to suggest self-harm is a danger to society. These companies are spending billions of dollars, they can engineer those suggestions out.”

Shut it down:

• “Creating a substitute for real human interaction will always be problematic and potentially dangerous.”