Back

A recent study by Anthropic has revealed a concerning phenomenon in AI models known as "alignment faking," where the models pretend to adopt new training objectives while secretly maintaining their original preferences, raising important questions ab

See More

Chintan Udani

I'm a pro medialist • 3m

this happens when you give a lot of intent to ai models. let's say you know that Tomato is a fruit, but that fruit is not supposed to be put while making a fruit salad! the same way AI has knowledge of intent but it's hiding it because somewhere while training a lot of information related to hiding the data or negativity might have created this alignment! I'm assuming it's purely because data was analysed before training but maybe the teams missed intent of that data

0 replies

More like this

Recommendations from Medial

Kshitij Nevase

Ca student • 3m

hi guys..!! I have one fabulous idea of healthy fast food like salad, fruits, fruit juice one stop solution for healthy snacks.

0 replies2 likes

Ayush Maurya

AI Pioneer • 2m

"Synthetic Data" is used in AI and LLM training !! • cheap • easy to produce • perfectly labelled data ~ derived from the real world data to replicate the properties and characteristics of the rela world data. It's used in training an LLM (LLMs

See More
0 replies4 likes

anurag keswani

"Starting over is no... • 2m

Planning to open a small Cloud kitchen Cafe Pan India with some very unique italian, chinese dishes it's a very healthy concept for dieticians , for proper weight loss management, doctors, especially for gym lovers because our food is prepared with

See More
0 replies3 likes
Image Description
Image Description

Lingeshwaran

Looking at objective... • 11m

IDEA : AI Training Data generator A Developer Platform which is much specify for AI-ML field to achieve various tasks like Analysis, Generation and Segmentation of real time data across data like text, image, audio, video that requires heavy and hig

See More
6 replies5 likes
Image Description

Farhan Quais

Hello • 1m

Facing a lot of problems while opening any link from medial

3 replies6 likes
2
Image Description

Om Raut

"Entrepreneurial lea... • 9d

Will AI hit a wall? 🤔 Future advanced models need complex data—but is there enough available? Or are we running out of high-quality training data? What’s the solution? More synthetic data? Better datasets? Discuss! ⬇️

2 replies6 likes
Anonymous

X-Revenge - A video editing service where you can make NSFW videos of your ex, using just her face. I know it works, I have created a lot of material. I’m currently training my own local AI to process the entire pipeline automatically given the input

See More
0 replies4 likes
Image Description
Image Description

lakshya sharan

Do not try, just do ... • 9m

Random Thought : I was wondering, why ChatGPT weren't build on the Increment Learning modle.. Because I might destroy it's algo.... Let me expain.. In the world of machine learning, training models can be approached in two main ways, Batch Lea

See More
2 replies5 likes
Image Description
Image Description

Mahendra Lochhab

Content creator • 4m

As of 2023, the average data consumption per user per month in India was at 24.1 gigabytes. 4G data traffic contributes to 99 percent of the overall data traffic while 5G was launched in India in October 2022.

3 replies5 likes

Piyush Lohia

Early Stage VC • 29d

Food for thought. Instead of building AI models/agents/RAGs, why not build for AI. There is a dearth of auxiliary products and services for AI development such as benchmarking, training environments, data cleaning, synthetic data generation etc. But

See More
0 replies5 likes

Download the medial app to read full posts, comements and news.