🚫 Call to Halt Superintelligent AI Development
Over 800 public figures, including AI pioneers Geoffrey Hinton and Yoshua Bengio, Steve Wozniak, and Prince Harry, have signed an open letter urging a ban on developing superintelligent AI until it’s proven safe. The letter cites risks from job loss to extinction, echoing public concern amid rapid AI advances.
👓️ Amazon unveils AI smart glasses for its delivery drivers
Amazon announced it is developing AI‑powered smart glasses designed for its delivery drivers. The glasses enable hands‑free tasks such as scanning packages, obtaining walking directions, and capturing proof of delivery while also using computer vision to detect hazards such as pets or low‑light conditions.
🚪 Meta Slashes 600 AI Jobs Amid Restructuring (Paywall)
Meta cut 600 roles from its Superintelligence Labs, trimming staff in its AI research, infrastructure, and product divisions. The layoffs follow years of over-hiring and aim to speed up decision-making. Meta is still aggressively hiring for its core superintelligence team, now led by ScaleAI co-founder Alexandr Wang.
🚘 GM to Roll Out Eyes‑Off Driving by 2028
General Motors announced new in‑car tech, including Google’s Gemini AI assistant launching in 2026 and an “eyes‑off” self‑driving system debuting on the Cadillac Escalade IQ in 2028. The automaker also plans a new computing platform, home energy systems, and expanded AI features across models.
Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.
📽️ VIDEO
New DeepSeek Just Did Something Crazy...
DeepSeek’s new OCR model compresses text 10× using images, boosting LLM context windows and efficiency—an insane leap in AI compression tech.
🌱 SUSTAINABILITY
AI’s Environmental Payback: 5 Ways It’s Helping Cut Emissions
Artificial intelligence is under scrutiny for its steep energy demands, data centers alone consumed 1.5% of global electricity in 2024, per the IEA, but scientists and startups are using AI to offset its footprint. From optimizing building energy use and electric vehicle (EV) charging to slashing methane emissions and improving geothermal drilling, AI is being deployed to reduce carbon output across sectors.
Google’s Project Green Light, now in 20 cities, fine-tunes traffic signals to curb vehicle emissions by up to 10%. Experts argue that these gains could outpace AI’s growing power consumption if scaled effectively. → Read the full article here.
🧪 LLMS
What Is AI Poisoning? Tiny Data Hacks Can Corrupt Large Language Models
“AI poisoning” — the deliberate insertion of corrupt or misleading data into an artificial intelligence (AI) model’s training or fine‑tuning process — can covertly compromise a model’s knowledge and behavior. Relatively small amounts of poisoned data or backdoor triggers can introduce targeted failures or steer behavior toward false outputs. This threat is more immediate than often assumed, and it raises serious concerns for trust, safety, and misinformation in deployed AI systems.
Highlights
“AI poisoning” refers to the attacker’s act of intentionally teaching an AI model incorrect or harmful information.
Two key forms are outlined: data poisoning (inserting malicious training data) and model poisoning (modifying the model after training).
A direct method, backdoor attacks, causes the model to behave normally until activated by a specific, often rare input.
Indirect manipulation includes topic steering, where attackers flood datasets with biased or false content to shift model outputs.
Replacing as little as 0.001% of training tokens with misinformation has led to models producing more harmful or inaccurate responses.
The analysis stresses that large language models (LLMs) are more fragile than they appear — small interventions can cause outsized effects.
Forward Future Takeaways
LLMs are not inherently robust — minimal tampering can trigger harmful behavior. As these systems scale, securing training pipelines and continuously monitoring for subtle manipulation must become standard practice. Without active defenses, poisoning risks may quietly escalate alongside model adoption. → Read the full article here.
📚 RESEARCH
LLMs Trained on Clickbait Show Measurable Declines in Reasoning
Researchers from Texas A&M, UT Austin, and Purdue tested the “LLM Brain Rot Hypothesis” by training large language models on junk data—specifically, clickbait content and viral posts from X. The study, published as a preprint on October 22, 2025, found that exposure to low-quality web content degraded model reasoning, context understanding, and safety adherence.
Meta’s Llama 3 was especially vulnerable, developing what researchers called “dark traits” like narcissism and psychopathy. Attempts to reverse the damage with mitigation techniques proved only partially effective, reinforcing calls for better data curation in model training. → Read the full article here.
🛰️ NEWS
What Else is Happening
🦿 Amazon Eyes 600K Job Cuts via Robots: Leaked docs show Amazon aims to replace 600,000 US roles by 2033 with automation, saving $12.6B, cutting 30¢ per item by 2027.
✨ GM Adds Gemini AI to Cars in 2026: GM will bring Google’s Gemini AI to OnStar-equipped vehicles as a voice assistant upgrade, enabling natural speech, web queries, and car-specific controls.
📈 Fal AI Hits $4B+ Valuation: The multimodal platform serves 2M+ developers building with image, video, audio, and 3D models. Just months ago, Fal closed a $1.5B Series C.
👮 AI Prank Triggers Police Raid: A Maryland woman used AI to fake a home invasion for a TikTok-inspired prank, prompting a major police response and landing her with criminal charges.
💰 Oculus Founders’ AI Startup Raises $250M: Sesame, building voice-driven AI smart glasses, secured $250M Series B and opened its iOS app beta, promising natural, expressive AI conversation.
⚡ Applied Digital Lands $5B AI Lease: Applied Digital signed a 15-year, $5B deal with a U.S. hyperscaler for 200 MW at its North Dakota site, expanding total leased capacity there to 600 MW.
🧰 TOOLBOX
Trending AI Tools
👨💻 Waver: Build and deploy custom vision-language models with open-source AI tools.
🗣️ Synthesys: Generate lifelike voiceovers and AI video avatars for marketing, e-learning, and media.
📲 App Alchemy: Design and launch AI-powered apps with no code using custom workflows and tools.
🏆️ REFERRALS
Share and Get 🔥 Prizes
🗒️ FEEDBACK
Help Us Get Better
What did you think of today's newsletter? |
That’s a Wrap!
❤️ Love Forward Future? Spread the word & earn rewards! Share your unique referral link with friends and colleagues to unlock exclusive Forward Future perks! 👉 Get your link here.
📢 Want to advertise with Forward Future? Reach 600K+ AI enthusiasts, tech leaders, and decision-makers. Let’s talk—just reply to this email.
🛰️ Want more Forward Future? Follow us on X for quick updates, subscribe to our YouTube for deep dives, or add us to your RSS Feed for seamless reading.
Thanks for reading today’s newsletter—see you next time!
Matthew Berman & The Forward Future Team
🧑🚀 🧑🚀 🧑🚀 🧑🚀
