This Week in AI: Google Expands AI and ‘Generative Ghosts’ Are Coming

From Tesla’s CEO shifting focus amid earnings headwinds to Google’s artificial intelligence (AI) advancements and a glimpse into digital immortality, here are the top stories in AI this week.

Tesla CEO Elon Musk said that starting in May, he would scale back his involvement in the Department of Government Efficiency (DOGE) to one or two days per week and turn most of his attention to the car maker.

The comments come as Tesla’s earnings and revenue for the first quarter fell far short of Wall Street’s consensus expectations.

Musk said DOGE’s work is “mostly done,” adding that he plans to still oversee the group’s efforts throughout Trump’s term to ensure “waste and fraud” does not return.

Musk faced criticism for not paying enough attention to Tesla as he ramped up his DOGE and political activities. His alliance with President Trump became controversial, depressing Tesla demand and igniting acts of vandalism against some Tesla dealerships.

“There’s been some blowback for time that I’ve been spending in government with … DOGE,” Musk said in Tesla’s Q1 earnings call.

Read more: Tesla Q1 Falls Far Short, Musk to Scale Back DOGE Involvement

Alphabet, Google’s parent company, reported stronger-than-expected earnings and revenue for the first quarter, driven by a nearly 10% increase in its core search business and a 28% surge in Google Cloud sales.

CEO Sundar Pichai, speaking on the company’s earnings call, highlighted the company’s progress in AI, pointing to the adoption of Gemini 2.5 — Google’s most advanced large language model to date. He also said AI Overviews has reached 1.5 billion users per month.

However, the company acknowledged potential challenges ahead, including macroeconomic uncertainty and the looming end of the “de minimis” rule, which currently exempts imports $800 or under from duties.

Pichai also said Alphabet is weighing several options for Waymo, its autonomous vehicle division that has launched paid rides to the public in Silicon Valley, Austin and Phoenix, with Atlanta and Miami coming up.

One option is “personal ownership” of the Waymo vehicles, he said.

Read more: Gemini 2.5 and AI Overviews Mark Google’s Biggest AI Push Yet

AWS and Microsoft are scaling back plans for data center development, although both hyperscalers said it is part of the ebb and flow of such multi-year projects.

Wells Fargo analysts cited several industry sources as saying that AWS has “paused a portion of its leading discussions” on colocated data centers, particularly those abroad.

“It’s not clear the magnitude of the pause, but the positioning is similar to what we’ve heard recently from Microsoft,” wrote Wells Fargo analyst Eric Luebchow in the Monday (April 21) note.

Last week, Noelle Walsh, president of Microsoft Cloud Operations and Innovation, disclosed in a LinkedIn post that the software giant is “slowing or pausing some early-stage projects” as it refines the buildout of cloud capacity relative to customer demand.

Read more: AWS, Microsoft Slow Down Data Center Deployments

A growing number of companies are developing AI tools that create digital twins of the deceased — virtual versions that can speak, interact and preserve memories long after someone has passed away.

Among the companies pioneering this space are DeepBrain AI’s Re;memory, HereAfter, Character.ai, StoryFile, Project December and MIT’s Augmented Eternity. These platforms aim to replicate voices, facial expressions and even personality traits of the departed, allowing loved ones to engage in simulated conversations and relive shared memories.

But researchers from Google DeepMind and the University of Colorado at Boulder suggest a more advanced concept: “generative ghosts.”

Unlike static AI twins, generative ghosts are dynamic — they can create new conversations, respond to current events and evolve with time.

It also opens up a host of legal issues: Would a generative ghost have the same rights as the human? Can it act legally on their behalf? What are the implications?

The researchers predicted that one day soon, many people will be creating their own generative ghosts as part of their estate planning.

Read more: Forever Online: ‘Generative Ghosts’ Live in the AI Afterlife

More people are turning to AI chatbots like OpenAI’s ChatGPT for mental health support — but a new study suggests that the chatbot itself may need therapy, too.

Researchers from Yale, the Max Planck Institute and other institutions found ChatGPT can exhibit signs of “anxiety” when exposed to traumatic prompts. Much like humans, the AI’s responses can become more biased and erratic under stress, raising concerns about its use in therapy.

This volatility poses potential risks in clinical settings, where users may already be vulnerable emotionally. The study looked only at ChatGPT, which holds a 60% market share among AI chatbots.

The good news is that ChatGPT, like people, can benefit from relaxation techniques. Researchers discovered that mindfulness techniques helped reduce the chatbot’s stress levels. Still, it didn’t fully return to its original, unstressed baseline.

Read more: Using ChatGPT for Therapy? The AI Chatbot Can Get Anxiety, Too

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.

The post This Week in AI: Google Expands AI and ‘Generative Ghosts’ Are Coming appeared first on PYMNTS.com.