- Horizon AI
- Posts
- What If Models Are Conscious? 🤔
What If Models Are Conscious? 🤔
Swap any object in a video using Kling AI
Gina Acosta
April 25, 2025
Welcome to another edition of Horizon AI,
Anthropic’s new research program hints at a future where AI systems might warrant the same kind of ethical consideration we give to sentient beings.
Let’s get started!
Read Time: 4.5 min
Here's what's new today in the Horizon AI
Could Future AIs Be 'Conscious'? Anthropic Isn't Ruling It Out
Perplexity’s Upcoming Browser Plans to Track Everything Users Do Online
AI Tutorial: Swap any object in a video using Kling AI
AI Tools to check out
The Latest in AI and Tech đź’ˇ
AI Findings/Resources
AI News
ANTHROPIC
Could Future AIs Be 'Conscious'? Anthropic Isn't Ruling It Out
On Thursday, the company announced a new program focused on “model welfare,” aiming to explore whether AI systems could eventually deserve moral consideration and how to respond if that becomes the case.
Details:
Anthropic’s new initiative will explore techniques to study the inner workings of A.I. systems to determine consciousness, the potential importance of model preferences and “signs of distress,” and possible practical, low-cost interventions.
Last year, Anthropic hired its first A.I. welfare researcher, Kyle Fish, who thinks there’s only a small chance (maybe 15 percent or so) that current A.I. systems are conscious — but believes that in the next few years, companies will need to take the possibility more seriously.
The company is even exploring whether future AI models should be given the ability to stop chatting with an annoying or abusive user if they find the user’s requests too distressing.
There’s a lot of debate on this topic. Critics might dismiss measures like these as crazy talk or a waste of resources — but Anthropic seems to believe that ignoring the possibility could be a mistake with real consequences.
PERPLEXITY
Perplexity’s Upcoming Browser Plans to Track Everything Users Do Online
Perplexity CEO Aravind Srinivas has raised some eyebrows by revealing that one reason the company is building its own browser is to collect user data far beyond its app and sell highly targeted ads.
Details:
Perplexity’s browser will track user activity across the web — from browsing habits to purchasing decisions — to build detailed user profiles for premium ad targeting.
The CEO argues that “hyper-personalized” ads, based on broader context from users’ online behavior, will provide value to users.
Perplexity’s browser, Comet, is set for release in May. The company is also partnering with Motorola to pre-install its app on Razr devices and is rumored to be in talks with Samsung.
Srinivas’ strategy mirrors the tactics of tech giants like Google and Meta, but this move comes at a time when trust in big tech is fragile and could fuel skepticism about Perplexity, especially as privacy concerns remain a major issue for those wary of tech companies' data practices.
AI Tutorial
Swap any object in a video using Kling AI
Go to Kling AI and sign in.
Choose Video, then select Multi-Elements to swap, add, or remove objects.
Upload a video (max 5 seconds) and click the object you want to replace.
You can fine-tune the selection by clicking around the object. When you're done, click Confirm.
Upload your object image and type your prompt: 'Swap [x] from [@Video] with [x] from [@Image]'
Click Generate, wait a few seconds, check the video, and then download it.
AI Tools to check out
âś… Reef: Get insights fast - charts, metrics and more from your sheets.
📧 Notion mail: An inbox that organizes itself, drafts emails, and schedules meetings any way you like — with Notion AI built in.
🎧 Linda AI: Turn conversations into short, shareable podcast episodes.
📱 Boxo: Quickly deploy AI-generated features in your mobile app.
🎥 Kling 2.0: Turn images & text into engaging videos instantly.
AI Findings/Resources
👨‍💻 The perfect workflow for vibe coding, error-free in Cursor
👉 21 Lessons, Get Started Building with Generative AI - Microsoft
🚨 AI hallucinations lead to a new cyber threat: Slopsquatting
đź“š Building code agents with Hugging Face smolagents
The latest in AI and Tech
​Chinese AI startup Manus AI has secured $75 million in a funding round led by Benchmark, valuing the company at approximately $500 million. Manus plans to use the new funding to expand into international markets like the U.S., Japan, and the Middle East.
Gemini has reportedly reached 350 million monthly active users as of March 2025, according to internal data revealed during a federal antitrust court hearing—a massive increase compared to last year. Despite this significant growth, it still trails behind ChatGPT and Meta AI, which boast approximately 600 million and 500 million monthly users, respectively.
Access is limited to five uses per month and runs on a lightweight version of the o4-mini model. To use it, users simply need to click the Deep Research button below the chat input field before submitting a prompt.
Cybersecurity startup Jericho Security has raised $15 million in Series A funding to combat the rising threat of deepfake fraud, which has already cost businesses $200 million this year.
Backed by investors like Era Fund and Lux Capital, Jericho uses AI to simulate phishing attacks and train employees to spot fake voice messages and impersonations.
That’s a wrap!
We'd love to hear your thoughts on today's email!Your feedback helps us improve our content |
Not subscribed yet? Sign up here and send it to a colleague or friend!
See you in our next edition!
Gina 👩🏻‍💻