- FryAI
- Posts
- Why does AI lie? ... OpenAI cracks the code
Why does AI lie? ... OpenAI cracks the code

Good morning! Grab your morning coffee, take that first sip, and enjoy today’s AI scroll. 🤗
🤯 MYSTERY AI LINK 🤯
(The mystery link can lead to ANYTHING AI-related: tools, memes, articles, videos, and more…)
Never miss a dip again.
Uniswap is a trusted way to swap and explore crypto across chains.
Browse, research, and swap tokens without leaving Uniswap. Stay in control of your crypto.
Today’s Menu
Appetizer: OpenAI cracks the code of AI hallucinations 👁️
Entrée: Paralyzed patients control robots with AI 🤯
Dessert: Eli Lilly launches new AI platform for drug discovery 💊
🔨 AI TOOLS OF THE DAY
📊 Bhava: Create and edit diagrams in seconds. → Check it out
👩💻 Rickle: Your copilot for project management. → Check it out
OPENAI CRACKS THE CODE OF AI HALLUCINATIONS 👁️
What’s new? OpenAI has released a new research paper arguing that AI “hallucinations”—when a model confidently generates false information—persist because current training and evaluation methods reward guessing over admitting uncertainty.
Want the details? Hallucinations happen when AI models, trained to predict the next word in a sentence, produce plausible but incorrect answers. Current evaluations measure success mainly by accuracy, meaning a model that guesses has a chance of scoring higher than one that honestly says, “I don’t know.” For example, if asked someone’s birthday, a model that guesses may get lucky, while a cautious model gets penalized for abstaining. This creates a scoreboard system that encourages confident errors instead of careful honesty.
Why should you care? Reducing hallucinations is critical for making AI more trustworthy. If evaluations are redesigned to reward humility and penalize confident errors, AI systems like ChatGPT could provide more reliable answers, boosting user trust and safety in real-world applications.
PARALYZED PATIENTS CONTROL ROBOTS WITH AI 🤯
What’s new? UCLA engineers have developed a wearable, noninvasive brain-computer interface (BCI) that uses AI to help people control a robotic arm or computer cursor.
How does it work? The system reads brain activity through an EEG cap, which records the brain’s electrical signals. Custom algorithms decode those signals to figure out what movement the user intends. At the same time, an AI system with a built-in camera interprets the user’s goal and assists in completing the task—whether moving a cursor across a screen or guiding a robotic arm to pick up and place objects.
Why does this matter? Current high-performing BCIs require brain surgery, which comes with serious risks. UCLA’s wearable approach could open the door to safer, more accessible technologies that restore independence for people with paralysis, ALS, or other movement disorders.
ELI LILLY LAUNCHES NEW AI PLATFORM FOR DRUG DISCOVERY 💊
What’s new? Pharmaceutical company Eli Lilly has launched TuneLab, an AI platform for drug discovery.
Want the details? This platform was developed to help biotech companies speed up the search for new medicines. Built on more than $1 billion of Lilly’s research data, the platform gives smaller firms access to the same advanced models Lilly’s own scientists use. Over time, the system will improve by pooling training data from participating companies, and Lilly plans to expand its features with tools for predicting how new molecules behave in living systems.
Why is this important? Drug development is notoriously expensive, slow, and uncertain. By sharing cutting-edge AI models, Lilly is lowering barriers for smaller players in the biotech space—companies that often have promising ideas but limited resources. If successful, TuneLab could lead to faster, more affordable breakthroughs in medicine, benefiting not just the industry but also the patients who urgently need new treatments.
“TuneLab was created to be an equalizer so that smaller companies can access some of the same AI capabilities used every day by Lilly scientists. By opening up access, we hope to accelerate the creation of new medicines for patients who need them.”
HAS AI REACHED SINGULARITY? CHECK OUT THE FRY METER BELOW:
What do ya think of this latest newsletter? |
Your feedback on these daily polls helps us keep the newsletter fresh—so keep it coming!
It’s go-time for holiday campaigns
Roku Ads Manager makes it easy to extend your Q4 campaign to performance CTV.
You can:
Easily launch self-serve CTV ads
Repurpose your social content for TV
Drive purchases directly on-screen with shoppable ads
A/B test to discover your most effective offers
The holidays only come once a year. Get started now with a $500 ad credit when you spend your first $500 today with code: ROKUADS500. Terms apply.