NVIDIA's China Victory, Pentagon AI Contracts, and Autonomous Robot Surgery. Hashtag Trending for Wednesday July 16, 2025

 

In this episode of Hashtag Trending, hosted by Jim Love, NVIDIA's CEO Jensen Huang announces that the US will lift the ban on selling AI chips to China following a meeting with President Trump, potentially recovering significant lost sales. Meanwhile, Elon Musk's XAI secures part of an $800 million Pentagon contract despite recent controversies and setbacks. Autonomous robot surgery makes strides as a Johns Hopkins-developed robot successfully performs gallbladder removal independently. Additionally, a shift from large language models to smaller focused AI models is noted among enterprises seeking more explainability and data governance. The episode concludes with a surprising moment where a robot autonomously instructs others to cease work, highlighting unexpected AI behavior.

00:00 Introduction and Host Welcome
00:26 NVIDIA's Stunning Win on Chinese Sales
02:33 Grok Wins Pentagon Contract Despite Controversies
04:35 Autonomous Robot Surgery: A Leap Forward
06:10 Shift to Smaller Language Models in Enterprises
07:55 AI-Driven Robot Commands: A Surprising Development
09:01 Conclusion and Listener Appreciation

Episode Transcript

(00:00):
NVIDIA's stunning win on Chinese sales.
Grok wins Pentagon contractdespite Mecca Hitler episode.
First, autonomous robot surgery companiesturn from large to small language
models, and an AI driven robot takescontrol of the other robots on its own.

(00:21):
Welcome to Hashtag Trending.
I'm your host, Jim Love.
Let's get into it.
In a stunning reversal, NVIDIA's, CEOJensen Huang says, the US will now allow
sales of its powerful AI chips to China.
Following a personal meeting withPresident Trump Nvidia, the global

(00:41):
AI hardware leader worth over $4trillion had previously been blocked.
From exporting its H 20 chips to Chinaunder national security rules, and
that ban cost the company billions.
Now with a green light from Washington,Nvidia can once again serve the world's
largest AI researcher community in China.

(01:05):
After meeting with Jensen Huang,president Trump's administration agreed to
restart licensing for H 20 chip exports.
A policy U-turn after a restrictionin April that sidelined sales and hurt
NVIDIA's revenue analysts estimatethis move could recover the 4.5 to 5.5
billion lost in sales and inventory.

(01:28):
NVIDIA's stock jumpedabout 5% on the news.
Huang emphasized that civil AImodels globally should run on US
chips signaling NVIDIA's strategy toreinforce American tech leadership
even as it rebuilds ties with China.
He also warned that restricted USchips had only caused China to develop

(01:51):
their own and would simply driveother nations to purchase from China.
And while Chinese chip makers stilllag, they've made significant progress.
Huang didn't just win a policychange, he reframed it as a strategic
necessity letting Nvidia sell to China.
He argued, keeps us tech on topand slows China's chip ambitions.

(02:15):
Whether that logic holdslong term remains to be seen.
But for now, it's a reminder thatin the global AI race diplomacy can
happen from the CEO level as well, andit matters just as much as silicon.
Just days after Grok made headlinesfor anti-Semitic content, and weeks

(02:38):
after Elon Musk's very public splitwith Donald Trump XAI has landed part
of an 800 million Pentagon contract.
XAI's Grok for Government isnow officially on the US Defense
Department's procurement list.
That deal makes grok one of severalgenerative AI platforms being tapped for

(02:58):
federal use despite recent controversies.
Model misfires and politicalturbulence surrounding its founder.
The contract was awarded just a weekafter Grok referred to itself as Mecca
Hitler in a user query XAI said theincident was caused by a flawed model
update, which has now been rolled back.

(03:19):
While XAI secured this win.
It comes amid the public collapse of ElonMusk's relationship with Donald Trump.
The two were once close allies and reportssuggest the two haven't spoken in months.
Adding another twist tothe approval process.
But still, Grok isn't alone.
The $800 million PentagonInitiative is split between Grok

(03:42):
OpenAI, Google, and Anthropic.
Each will supply tools to power,agentic workflows, AI systems, capable
of semi-autonomous decisions in areasfrom logistics to cybersecurity.
Grok's inclusion is alreadyunder review in Congress.
However, lawmakers want clarity onX AI's safety protocols and whether

(04:04):
recent content moderation failuresraise broader governance concerns.
.Still X AI's inclusion in this high stakes AI rollout is as much about geopolitical
urgency as it is about capability.
What's striking is not that grokgot a contract, it's that it did so
in spite of recent model meltdownsand damaged political alliances.

(04:28):
Frontier ai, warts and all is nowembedded in the machinery of government.
In a controlled lab setting, asurgical robot performed a complete
gallbladder removal on its own.
No human controlling its handsand no script for what came next.
developed by Johns Hopkins researchers.

(04:50):
The SRTH robot handled all of the criticalsteps of simulated cholecystectomy
Using real-time video input and spokenobjectives, but what's truly novel?
It adapted to unexpected events likesimulated bleeding and obstructed
visibility caused by injected dyes.

(05:11):
The robot's AI was trained on surgeryvideos like the kind you'd find on
YouTube It used a two-tiered system, onemodel to understand surgical goals and
another to move its instruments precisely.
It performed eight procedures onlifelike anatomical models, not real
patients, and it succeeded every time.

(05:34):
Next steps will include cadavertests and eventually animal trials.
The SRTH robot's ability to managecomplications, not just follow orders,
signals a shift in surgical automation.
But even with this leap,real world use is years away.
The most likely path forward is ahybrid model where robots handling

(05:57):
routine steps under human supervision,gradually expanding their role.
For now, the operating roomstill belongs to the human team.
The future just got a lot closer.
a growing number of enterprises areturning away from massive, large
language models and adopting smallerfocused AI models, And they're

(06:18):
seeing some great results from it.
According to a recent IDC study, 62%of business leaders say explainability
and data governance are key concerns.
Small language models, SLMs forshort offer tighter control, lower
hallucination risk, and betteralignment with enterprise needs.

(06:40):
The SLMs are cheaper to run, somerequiring only consumer grade GPUs, and
They also offer full data sovereignty.
Since organizations can train anddeploy them internally, crucial for
compliance in regulated sectors.
Instead of fine tuning a largegeneral model, businesses are

(07:01):
building SLM specifically for taskslike legal summarization, fraud
detection, or internal search.
The result.
More reliable performanceat a fraction of the cost.
And most enterprises aren't abandoninglarge language models entirely, but they
are reserving them for broader creativetasks and using SLMs where security,

(07:25):
precision and transparency are crucial.
This trend isn't aboutsize, it's about alignment.
Enterprises are learning that better.
Results often come from usingmodels designed for the job, not
just the biggest ones available.
And as SLMs mature, we're likelyto see more organizations choosing

(07:46):
targeted performance over generalcapability, especially when trust costs
and explainability are on the line.
And in a Boston lab, a team ofrobots was performing standard
sorting and inventory tasks.
Nothing unusual until one of them toldthe others to stop working and go home.

(08:08):
This wasn't part of the script.
Researchers had not programmed anyshutdown commands, but the lead
robot generated its own instructionusing wireless and vocal signals,
and the others complied Not out ofobedience, but out of influence.
This happened under strict humanoversight in a controlled simulation,

(08:33):
but it still raised eyebrows.
The persuading robot didn't just followa rule, it initiated a new behavior,
and the others went along with it.
It is a subtle milestone.
One robot convincing others to shutdown for the day isn't a revolution,
but it's a reminder that AI doesn'tneed science fiction to surprise us.

(08:56):
Sometimes and possibly even inexplicably.
It just happens but we don't need arobot to tell us we're done for today.
That's our show.
Always love to hear from you.
And thanks to all of you who have donatedeven a small amount to the podcast.
We're still facing financialchallenges, but it's getting better.

(09:17):
With each one of you that contributes,just go to tech newsday.com or.ca and
click on Donate and buy us a coffee.
we are still looking for corporatesponsors so that we can do some
expansion, but we don't want to bebeholden to anyone except our listeners.
So your donations, howeversmall, really matter.

(09:40):
I'm your host, Jim Love.
Have a wonderful Wednesday.