• At HumanX in Las Vegas tech leaders talked about on AI's impact on business.
• Speakers like Kamala Harris emphasized collaboration between the tech industry and government.
• AI enthusiasm ran high but many attendees were focused on practical ways to drive company revenue.

This week, I attended HumanX, where thousands of startup founders, investors, and tech industry executives gathered in Las Vegas to discuss AI's value proposition to their bottom lines — and the future.

The nearly four-day conference opened with remarks from former Vice President Kamala Harris. Harris, who served as "AI czar" during the Biden Administration, was one among a handful of lawmakers at the event who called for stronger cooperation between the tech industry and the federal government.

In the following days, attendees sat in on panels with veteran investors like Vinod Khosla and Tim Draper alongside rising stars of the AI boom like Hugging Face's Thomas Wolf and Mistral's Arthur Mensch. There was ample time to swerve between the booths on the convention floor where tech giants and startups alike had reps ready to woo customers.

AI conferences can feel a bit like a dance. People talk with excitement about the transformative power of this technology, but it's also clear they're gauging where the real value lies.

Here are my top takeaways:

1. Patience is a virtue when it comes to ROI

Conference speakers had some advice on making long-term gains with AI.

Glean CEO Arvind Jain cautioned companies to establish an "AI team" to create a roadmap for integrating the technology. "Make small bets. Don't try to actually immediately focus on ROI," he said. "Focus on education first, like, make your workforce trained with AI."

In the sales landscape, Conviction's Sarah Guo said companies have already made "high-fructose corn syrup" gains, reveling in their capacity to send mass emails or spam customers with calls.

"That actually drove a lot of traction," she said. However, "real customers actually churned off that pretty quickly because it doesn't serve a real need. Nobody wants the spam — which is what it is."

Still, it got people to start "buying the products" and she expects that to continue. "That's what I think we're going to see this year in terms of those real ROI use cases."

2. Vibes and valuations

AI startups are fetching wild valuations, but the question is whether they're caught in a bubble that's set to burst.

Investors are essentially "taking a bet" on the future cash flow of a business, said Tuhin Srivastava, CEO and cofounder of AI inference company Baseten. "I'm a  beneficiary of that," he added.

Srivastava pointed to Anysphere, which makes the coding assistant Cursor, and was in talks to raise funding at a valuation of close to $10 billion earlier this month, after hitting $100 million in annualized recurring revenue in a span of 12 months, Bloomberg reported. Coding startup Codeium, is raising funding at a $2.85 billion valuation after hitting $40 million in ARR, TechCrunch reported.

At face value, these seem like massive valuations, but Srivastava says he thinks they're reasonable in the context of the venture landscape just a few years ago.

"Companies in 2021, during all that craze were like, you know, a million dollars of ARR, raising a billion dollars. Think about these AI companies, they all have a lot of revenue," he said.

He joked that valuations — to some extent — are being calculated on "vibes" but he believes they're grounded in real growth potential. "You know, we're probably at 0.1 percent adoption of AI in the enterprise. So we have a thousand x upside there."

For companies with little to no revenue, talent may be the value metric.

A few investors told BI they heard, to their surprise, that companies acquiring AI startups for a strategic advantage were valuing them on a "per engineer" basis.

If you have more thoughts on talent-based valuations contact me at [email protected].

3. Here's at least one new job we'll see soon

Aside from prompt engineers and a beleaguered new set of "AI heads" at companies, many are still wondering what "new jobs" this technological shift will bring.

As AI makes it easier to code, and "vibe-code," we should expect to see a new wave of highly compensated "product engineers," Andrew Filev, CEO of AI coding platform Zencoder said. These are individuals who are versed in product management, a role that typically oversees new products from ideation to launch, and software engineering, which deals with the technical details of new products.

4. Governance is a pain point

AI governance is still a bit of a murky term.

"There's a little bit of a conflation of governance and regulation," said Navrina Singh, founder and CEO of AI governance platform Credo.

Singh told BI she defines governance with three questions:

1. Do you have a handle or understanding of risk?
2. How do you actually mitigate the risk of these technologies?
3. How do you future-proof your AI investments for potential policy changes — not only at the company level, but at the government level?

The most misinformed opinion companies have is that governance will slow them down in adopting AI. Credo's data has shown the opposite. " We are finding that companies are getting better ROI," she said.

Dataiku's CEO Florian Douetteau observed a similar anxiety around governance at an "executive field trip" the company held for the CEOs of its customer base in September 2024. The company further investigated CEOs' top AI anxieties and published them in a survey. Governance ranked high on the list with 94% of about 500 CEOs surveyed saying they suspect employees are secretly using generative AI tools without official approval.

5. More people are sounding the alarm on deflation

Silicon Valley leaders from Khosla to Sam Altman have expressed concerns that AI will spur deflation. Those fears were echoed by a handful of attendees at HumanX.

To be clear, the US is still experiencing inflation with the consumer price index rising about 2.8% for the twelve month period ending in February 2025, according to the US Bureau of Labor Statistics. However, technological shifts are often correlated with deflation, as they boost productivity and lower production costs.

"There is no denying that AI-based technologies are evolving rapidly and being adopted by people and enterprises," said Steve Berg, a partner at Lytical Ventures. "There are inflationary impacts happening as well, which offset the deflationary impacts of technology, but what happens when one side or the other becomes dominant?"

• The Chinese startup DeepSeek shook the tech world and markets when it released R1, its new AI model.
• The West is now trying to reproduce R1 on its own terms and cut out Chinese servers.
• Recreating R1 from scratch can help researchers build better models and validate DeepSeek's claims.

Silicon Valley doesn't want to get caught out again. It's scrambling to replicate DeepSeek's AI model, the cheaper Chinese tech that shook Wall Street and is freely available for anyone to adopt.

Companies like Microsoft and Amazon have already made versions of DeepSeek's R1 models available on their cloud platforms. This allows people to use the models, which appear to match the capabilities of models from rivals like OpenAI, while keeping data from being sent to servers in China.

But there are also attempts to replicate DeepSeek's cost-efficient AI from the ground up — and see whether all the Chinese AI lab's market-moving claims hold up.

One major effort is being led by Hugging Face, a platform on which researchers in artificial intelligence's open-source community can collaborate and share their trade-research notes and ideas for free.

Leandro von Werra, the head of research at Hugging Face, told Business Insider that the company expected to complete its replication efforts within weeks. He described the mood at Hugging Face as "kind of like, 'Avengers assemble,'" as they dissect the inner workings of R1.

DeepSeek obtained open-source licensing for its model from MIT, which means a lot of the vital components of the recipe needed to build R1 have been laid out in the company's publicly available technical paper.

However, there are some elements of R1 that remain unclear.

In a December paper on V3, DeepSeek's earlier model, the Chinese company said the training cost $5.6 million in total. The cost was calculated based on its use of H800 graphics processing units, a less powerful version of Nvidia's top chips, at a rental price of $2 per GPU hour.

Right now, no one can be quite sure what the actual development cost of R1 was, von Werra told BI.

DeepSeek's research paper also did not share what was required to bake reasoning capabilities into V3 to then produce R1.

That said, von Werra thinks it won't remain a mystery for long. "I don't know about the compute number. We can only guess at this time," he said. "I think one thing that's exciting about our reproduction is we're going to find out pretty quickly if the numbers hold up."

Learning from DeepSeek

Some corners of Silicon Valley responded swiftly to the launch of DeepSeek. This week, Meta set up "war rooms" for its researchers to analyze DeepSeek, The Information reported. Sam Altman, the CEO of OpenAI, said Tuesday his company would accelerate the release of "better models."

DeepSeek's decision to publish its findings and make its R1 model open gives researchers worldwide insight into its novel approach.

The main technique used to make R1 so capable was "pure reinforcement learning," DeepSeek's paper said. This, Hugging Face researchers said in a blog on Tuesday, can "teach a base language model how to reason without any human supervision."

The researchers also know more specific technical details about why R1 caused such a stir in Silicon Valley and wiped $1 trillion from US stocks on Monday. For instance, the reasoning model is what's known as a "mixture of experts" model — industry-speak for a model that can be "pre-trained with far less compute." It also involves subtle changes to its architecture by introducing techniques like "multitoken prediction," first introduced by Meta, that make models more efficient.

Hugging Face's von Werra said details like this from DeepSeek had helped the industry better understand how a reasoning model like OpenAI's closed-source o1 was built. "Everybody thought this is the secret that is going to take awhile to crack," he said.

DeepSeek comes to platforms

Days after its launch, DeepSeek flew to the top of Apple's top free apps chart. While everyone wanted to try the latest AI tool, DeepSeek's policy of storing user data in China prompted security concerns.

This spurred US companies to make R1 available on their own platforms so customers could use the Chinese AI model while cutting out China's servers.

Lin Qiao, the CEO of Fireworks AI and former head of the PyTorch team at Meta, told BI that one clear reason for doing so was to ensure AI developers and users continued to get access to top model innovations.

"The approach we have been taking is always to enable state-of-the-art models for developers the fastest," she said. "DeepSeek is one example."

Her company, founded in 2022, made R1 available on its platform after congratulating DeepSeek for "pushing the boundaries of what's possible in open models." It has been made available via its serverless service, as well as through on demand and for enterprise customers.

Others have followed suit. On Wednesday, Microsoft announced that it was making R1 available in its model catalog on its AI development platform Azure AI Foundry to make it "accessible on a trusted, scalable, and enterprise-ready platform."

Asha Sharma, a corporate vice president at the tech giant, wrote in a blog that R1 "offers a powerful, cost-efficient model" but one that it had done "rigorous red teaming and safety evaluations" on before introducing as a model to its library.

Amazon Web Services is making a similar move. Swami Sivasubramanian, the vice president of AI and data at AWS, said this week that the company's "commitment to AI accessibility" meant R1 was being made available on its platforms such as SageMaker and Bedrock.

Perplexity CEO Aravind Srinivas, whose AI startup made R1 available to users of its search platform this week, said Tuesday in an X post that downloading the model onto its servers could also help control the way it responds to user queries and ensure those queries don't go to servers in China.

DeepSeek's AI appears to censor sensitive information about China, such as refusing to answer questions about the 1989 Tiananmen Square protests. Srinivas said that Perplexity's version of R1 had no censorship and shared its accurate response to what happened in Tiananmen Square.

David Sacks, the White House's AI czar, offered one reason Perplexity's R1 integration was an important way to reproduce R1 in the West. "This is one of several ways that you can try DeepSeek R1 without downloading the app or sharing any data with a Chinese company," he said on X.

Many see DeepSeek as an example of China challenging American AI hegemony using a tried-and-tested playbook. OpenAI, which has a lot to lose from DeepSeek making its technology freely available, said on Wednesday it's investigating whether the Chinese firm "inappropriately" replicated its models for training.

For von Werra, it's a full-circle moment. The whole field started as open source, so seeing efforts to make a leading reasoning model available for free is welcome, he said.

"I think in the end, everybody's going to get better models and do cooler things," he said. "I feel like it's a win-win situation."

• AI's significant impact on workforce and tech use will continue in 2025, leaders say.
• Shifts in work dynamics, including the introduction of AI agents, are reportedly on the horizon.
• Tariffs under Trump may lead to higher prices and project cuts for businesses, Rimini CFO says.

Talks of artificial intelligence dominated 2024, and tech leaders predict that the next stage of the new AI era will come in 2025 — for better or for worse.

Their mixed predictions indicate that the outlook for AI in 2025 is still uncertain. Still, many execs assert that the way people interact with technology will continue to change in 2025, and it'll likely impact jobs.

Sam Altman, CEO of OpenAI, said in a Sunday blogpost that AI agents could join the workforce this year and "materially change the output of companies." Since generative AI became a more mainstream concept, concerns about job security have risen amongst workers.

Agentic AI, which requires GenAI to work, is the technology that AI companies say could work alongside their employees as digital coworkers.

"We continue to believe that iteratively putting great tools in the hands of people leads to great, broadly-distributed outcomes," Altman wrote.

This year will also reveal if the prediction that Bill Gates has been vocal about for over 10 years will come to fruition. Gates has said on many occasions that two-thirds of all jobs in the US will require some form of education beyond high school by 2025.

Here's what Amazon's CTO, an ADP exec, and other tech leaders say 2025 has in store.

The workforce is evolving.

Management services company ADP uses AI to assist sales reps, and GenAI helps the company prepare for investor days, among other things, according to The Wall Street Journal. Such investments will continue in 2025, Don McGuire, CFO at ADP, said to the Journal.

"Things that people used to sit beside you and have a headset, now you can do those things with GenAI tools," McGuire said.

More intentional tech use is "reshaping our relationship with the digital world," and people are prioritizing wellbeing over attention-seeking, Amazon CTO Werner Vogels said in a December blogpost.

"The workforce of tomorrow will not only be driven by financial success and career progression but by a deeper desire to create positive change in the world," according to Vogels.

As smarter tech integrates into the workforce, Vogels said "a quiet revolution" is happening among workers who value meaningful societal impact over financial success.

Millennials and Gen Z are leading the charge in finding their purpose at work, but Vogels said the trend is being driven by other age groups and the job market itself.

"Harnessing technology for good has become both an ethical imperative and a profitable endeavor," Vogels said.

However, Clement Delangue, CEO of AI startup Hugging Face, predicts a more combative response to AI. He said on LinkedIn that the "first major public protest related to AI" is coming in 2025.

Consumers will move away from distracting technology.

Tech users will find more intentional ways to use their devices in 2025, according to Vogels. The Amazon exec pointed at data related to social media use and mental health issues among teens in the US in his blog about 2025 as an indicator of a need to rethink our relationship with technology.

"Every swipe, headline, and notification are meticulously engineered to hook us," Vogels said.

More people are becoming conscientious of their screen time. Some are implementing rules for themselves or their children, and others are seeking out alternatives like "dumb phones" with no web browsers.

On the other hand, Delangue predicts that other consumers will buy into more futuristic devices, like the robots being developed by Tesla and other robotics companies.

"At least 100,000 personal AI robots will be pre-ordered," he said on LinkedIn.

Tariffs will bring about project cuts.

Donald Trump is weeks away from his second presidential inauguration, and the president-elect has already threatened tariffs on imported goods. It "feels like an inevitability," according to Michael Perica, CFO of software company Rimini Street.

LendingTree economist Jacob Channel previously told Business Insider that consumers will likely be subjected to higher prices for their goods during his term if Trump makes good on his pledge.

Companies may have to make some cuts to offset the effects of tariffs, Perica told the Journal. More businesses will have to adapt their strategies in anticipation of supply-chain disruptions.

"We are absolutely partnering with folks to help them along and take a look and evaluate what's a 'nice-to-have' project versus 'got to have,'" Perica said.

• Hugging Face's CEO predicts the first major AI protest and market disruptions in 2025.
• Clement Delangue also said AI advancements could halve a major company's market cap.
• His other predictions include China leading the AI race, driven by open-source model developments.

While people are preparing their New Year's resolutions, one AI company CEO has a different habit: locking in his predictions for what will happen in the industry in 2025.

Clement Delangue, CEO of the $4.5 billion startup Hugging Face, laid out six predictions for AI in the new year. He also scored himself on his last batch of predictions, which you can check out on LinkedIn.

This time around, Delangue expects major public backlash over artificial intelligence, sizable orders of personal AI robots, and China overtaking the US in the AI race.

You can take a look at his six predictions below.

The first major public protest against AI

While companies may be scrambling to incorporate AI innovations, not everyone is as eager for the AI era — and Delangue predicts they will be a lot more vocal next year.

"There will be the first major public protest related to AI," Delangue said in his post.

From professors struggling to combat rising plagiarism to AI-generated art controversies, artificial intelligence has led to frustrations and the uncertainty of change, which often leads to backlash.

AI will cut a big company's value in half

Describing what would basically be a CEO's nightmare scenario, Delangue also said that a large company could "see its market cap divided by two or more because of AI."

AI advancements could cause a major company's core technology or corporate value to become defunct, like how streaming impacted the DVD market.

In a reply to Delangue's post, one LinkedIn user pointed out Teleperformance as a possible example. The call center company sank to a seven-year low in February, with shares dropping as much as 29%, due to concerns over AI disruption. A day earlier, Klarna had announced that its AI assistant could account for two-thirds of its customer service chats.

Personal AI robots

With companies including Tesla and Jeff Bezos-backed Physical Intelligence already developing AI robots, Delangue predicts that these robot assistants will soon be available in the mass market.

"At least 100,000 personal AI robots will be pre-ordered," he said.

Elon Musk, who has admitted he tends to be optimistic about timelines, has said Optimus robot has a "good chance" of some units being shipped out in 2025, he said in a Tesla earnings call. At an estimated cost of $20,000 to $30,000, the robots would likely remain a luxury item until the cost could be brought down.

In November, Agility Robotics was able to "employ" its robot Digit at GXO Logistics' Spanx womenswear factories. CEO Peggy Johnson previously told Business Insider that having robots perform tasks at home, like folding laundry, may take longer to develop.

"A household is a very chaotic environment: At any given moment, a child's ball runs across the room, and dogs run by," she said. "There's things that are in the way."

China will lead the AI race

With the US and China battling it out for AI dominance, Delangue predicts that "China will start to lead the AI race."

The Hugging Face CEO said that China's ascendance will be a "consequence of leading the open-source AI race."

Although Chinese competitors like ByteDance and Baidu have developed closed-source models, a few startups like 01.AI and Alibaba's Qwen2 have decided to make their models open-source.

In Hugging Face's June LLM Leaderboard, which compares the performance of major open-sourced LLMs, Alibaba's Qwen2 topped the ranking ahead of Meta's Llama 3 and Microsoft's Phi-3.

Breakthroughs in biology and chemistry

While AI is quickly percolating many industries, Delangue predicted biology and chemistry are two fields that will see "big breakthroughs."

In October, Google's DeepMind CEO Demis Hassabis and director John Jumper received a Nobel Prize in chemistry for their use of AI to predict protein structures with the DeepMind tool AlphaFold.

Earlier this year, Hassabis predicted AI-designed prescription drugs could enter clinical trials in the coming years.

"I would say we're a couple of years away from having the first truly AI-designed drugs for a major disease, cardiovascular, cancer," he said.

Economic and employment growth from AI

Delangue's final prediction is that the "economic and employment growth potential of AI" will begin to show itself in 2025.

For Hugging Face, in particular, he predicted that 15 million AI builders would be seen on the platform.

Despite the company failing to reach last year's prediction of 10 million, instead landing at 7 million, Delangue remains optimistic that AI builders will continue to grow.