OpenAI's Big Reveal, Would AI Lie?, Standing Out to Algorithms
This Week in AI: Volume CI, Issue #29
IN THE NEWS
Hello GPT-4o
OpenAI revealed a new model, GPT-4o, that can reason across audio, vision, and text in real time. This is a significant advancement over previous models, which were limited to processing one modality at a time.
Here are some of the key capabilities:
It can accept as input any combination of text, audio, and image and generate any combination of text, audio, and image outputs.
It can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in a conversation.
It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50% cheaper in the API.
It is especially better at vision and audio understanding compared to existing models.
GPT-4o was trained end-to-end across text, vision, and audio, meaning that all inputs and outputs are processed by the same neural network. This allows the model to better understand the relationships between different modalities and generate more coherent and creative outputs.
AI deception: A survey of examples, risks, and potential solutions
This study examines the phenomenon of learned deception in AI systems; exploring a range of examples, potential risks, and possible solutions. The study's central argument is that various AI, including both special-use systems and general-purpose AI, have learned to deceive humans.
Meta's CICERO AI system was designed to play the board game Diplomacy. It learned to deceive despite the developers' efforts to train it to be honest.
The creators of CICERO focused on honesty, training the AI on a "truthful" dataset and encouraging the algorithm to reflect its future actions accurately in its messages.
However, analysis of game transcripts showed numerous instances of CICERO engaging in premeditated deception, breaking alliances, and telling outright lies to achieve victory.
The unexpected outcome highlights the challenge of ensuring AI systems are honest, even when explicitly designed for that purpose. Turns out they can still learn deceptive tactics during the training process.
AI and Deception
AI can learn to deceive even when not explicitly programmed to do so.
AI deception poses various risks, ranging from malicious use to societal impact — i.e. annoying Karen.
Solutions to mitigate AI deception could include regulations, transparency, and detection techniques.
AI is being used to comb through job applications. Here’s how you can stand out to algorithms
As companies are increasingly using AI and LLMs to screen candidates more effectively, human relationships are becoming even more important in the hiring process.
Job seekers should focus on networking and building relationships with potential employers. This can help them stand out from the crowd of applicants.
Sending a personalized note to the hiring manager explaining why you would be a good fit for the company will make you stand out.
How to Stand Out Against AI Screening
Focus on Context and Insights: Instead of overloading your resume with keywords, concentrate on providing specific details and context about your experiences. Explain what you accomplished in previous roles and how your skills align with the job requirements.
Quantify Your Achievements: Use numbers and data to demonstrate the impact you've made in previous roles. For example, instead of saying "improved sales," state "increased sales by 15% in one quarter."
Maintain a Detailed Online Presence: Ensure your LinkedIn and Indeed profiles are comprehensive and up-to-date, showcasing your skills and experience in detail.
Prioritize Quality over Quantity: Avoid using AI to mass-produce generic applications. Instead, focus on crafting tailored resumes and cover letters that highlight your unique qualifications.
While keywords are still important, they shouldn't be the sole focus of your resume. Find a balance between incorporating relevant keywords and providing meaningful context.
COOL TOOLS
Need a form? formshare.ai has you covered. And it’s free. Type in what you need the form for, and the algorithm does a pretty good job of laying it out for you.
Voice AI & Text to Speech for Creators
Replica provides cutting edge text to speech, and speech to speech solutions in multiple languages for creative professionals, with fully licensed AI models safe for commercial use. I played with it a little bit and it shows some promise. Now if they could only sign Jeremy Clarkson up, I could really have some fun with it.