Welcome to our weekly roundup of premium handcrafted AI information.
This week AI voice clones received banned however grew to become tremendous sneaky.
Superior AI added to anxious artists’ angst.
And Indonesian politicians introduced a useless dictator again to life.
Let’s dig in.
Misaligned
The struggle to ensure AI behaves higher is ongoing. The FCC declared that AI-generated voices in robocalls are unlawful, so ‘Joe Biden’ most likely received’t be calling you once more.
Voice cloning is getting so good it’s arduous to inform who we’re chatting with anymore. IBM Safety made a really sneaky proof of idea to indicate how AI can hijack a key a part of the audio in a dialog. That is fairly scary.
Does AI show racial and gender bias when evaluating photos? Researchers discovered that it typically does, however they surprise if it’s all the time a nasty factor.
If we did finally make a superbly aligned AI that was utterly ‘protected’, what would that be like? Meet Goody-2, the world’s most accountable AI mannequin. I’m offended.
by way of GIPHY
Get inventive
Can AI be inventive? Is AI-generated artwork actually artwork? Sam takes a deep dive into how artists are below fireplace and investigates the affect of AI on creativity and jobs within the sector.
This Sunday, greater than a 3rd of the US inhabitants watched the Tremendous Bowl. Numerous these viewers tuned in purely for the commercials. AI dominated the Tremendous Bowl adverts and gave us some perception into how people really feel in regards to the tech.
McAfee launched its second “Fashionable Love” research, investigating the interplay between AI and romance. Sure, that’s a factor now. Customers of on-line courting platforms are getting inventive and utilizing AI to juice their profiles and rip-off individuals.
Is the scale of your dataset making you are feeling a bit insufficient? MIT researchers revealed an enchanting research that reveals that symmetry of knowledge might resolve small dataset woes.
Chasing chips
There doesn’t appear to be any finish to the insatiable demand for high-end AI chips. OpenAI CEO Sam Altman needs to make his personal and says he wants trillions (with a t) for his outlandish AI chip mission.
Whereas Altman goals of his personal AI chip factories, NVIDIA CEO Jensen Huang is advising international locations to construct their very own sovereign AI belongings. And he is aware of simply the place they might purchase the GPUs they’ll want.
We could not must pay for cloud computing or API calls quickly. NVIDIA has launched Chat with RTX, a demo that lets you run an LLM like Llama 2 or Mistral regionally in your Home windows PC utilizing one among its RTX GPUs.
AI throughout the globe
Microsoft and OpenAI found that quite a few high-profile cybercrime teams have been utilizing their AI instruments and promptly blocked their accounts.
They are saying these teams are backed by nation-states like North Korea, Iran, Russia, and China. Even the names of a few of these cybercrime teams sound a bit phishy. One group known as Salmon Storm. I want I used to be making this up.
I had ChatGPT make a pic they might use on their t-shirts for his or her subsequent cybercrime workplace team-building day. Simply kidding guys. Please don’t hack me.
Geopolitical tensions are influencing the place AI funding flows and who you are able to do enterprise with. Abu Dhabi-based AI agency G42 has lower ties with Chinese language corporations after coming below stress from the US authorities and traders.
If you happen to’re making an attempt to make sense of how you can use AI to market your enterprise globally, then join the upcoming MarTech Summit occasions that may cowl the APAC and EMEA areas.
In the meantime, China denies any nefarious intentions with its pursuit of AI tech. A number of days in the past Chinese language researchers unveiled a robotic toddler named “Tong Tong” which they are saying is as clever as a toddler aged between three and 4.
China says it needs to mass-produce humanoid robots by 2025. Thousands and thousands of robotic toddlers, what might presumably go fallacious?
The King’s Cross
Followers of British royalty received a dose of AI fakery this week. AI-generated books are flooding Amazon’s cabinets to take a position about King Charles’s most cancers prognosis. The parents at Buckingham Palace will not be amused.
AI surveillance expertise was utilized in a year-long trial at a London Underground station with some promising outcomes. The tech is prone to be rolled out wider however the line between security and Large Brother surveillance is a positive one.
Jailbird and Zombie politicians
Politics is all the time entertaining however this week it received further bizarre.
Ex-Pakistan Prime Minister Imran Khan is in jail however his celebration was nonetheless contesting the latest elections. He simply declared that he received the elections utilizing an AI avatar and a clone of his voice.
The US could have among the oldest presidential candidates however, to not be outdone, an Indonesian celebration used AI to resurrect a useless dictator for election propaganda. Former President Suharto appeared fairly good on this video, contemplating he’s been useless for 16 years.
Among the finest cartoons and posters in historical past have been impressed by political satire or propaganda. However due to the chance of hyper-realistic AI fakes, Midjourney is contemplating snubbing out AI-generated photos of Trump or Biden because the election looms.
The AI physician will see you now
You may spend a couple of hundred {dollars} an hour on a therapist, or you would simply purchase an Apple Imaginative and prescient Professional. Researchers developed an immersive AI remedy system named XAIA which mixes GPT-4 with Apple’s new VR headset.
Individuals within the trial gave optimistic suggestions on the expertise. It is also that they have been simply joyful to get to play with an Apple Imaginative and prescient Professional with out the $3,500 price ticket.
Apple Imaginative and prescient Professional pic.twitter.com/06KO0b74Er
— Aleksandr Morozov (@almormd) June 5, 2023
Discovering the appropriate sufferers to take part in a medical trial includes wading by means of 1000’s of pages of medical notes and data. These researchers discovered a method to make use of GPT-4V to do that quicker, higher, and lots cheaper than people can.
In different information…
Listed below are another clickworthy AI tales we loved this week:
And that’s a wrap.
After seeing IBM Safety’s voice clone hack I feel we’ll all must develop into tremendous skeptical actual quick.
How do you are feeling about bringing again useless presidents to assist Biden and Trump marketing campaign? Who’d you wish to see in motion? I’m considering, Andrew Johnson after a couple of drinks debating Richard Nixon. I’d watch that.
Altman asking for trillions and everybody nodding in settlement as an alternative of calling the nuthouse to make a reservation for him sums up how loopy the AI scene has develop into over the past 12 months.
We are able to’t wait to see what occurs subsequent. And when it does, we’ll let you understand proper right here.