#ai #jobs #futurist
Elon Musk and Anthropic’s deal to use SpaceX’s computing power is big news. Anthropic, the maker of Mythos AI, which is thought as a threat to world cybersecurity, should be kept within the boundaries of safe AI. It must be ensured that SpaceX’s Colossus 1 facility in Tennessee, which would be provided to Anthropic, is not used to develop unsafe AI.
This computing power of 220,000 Nvidia GPUs would power Anthropic AI engines. This is a huge compute power and must be used wisely.
We must first understand what safe and unsafe AI are before we can consider what the future of such a deal could be.
Safe AI includes narrow AI, which depends on a computer science algorithm to work. Safe AI can include other AI techniques, such as those for cancer detection. Safe AI can be classification algorithms that are white-box; these procedures’ outputs can be predicted well in advance. These can include clustering, classification, summarization, retrieval, and categorization with safe algorithms.
Unsafe AI can be made using deep learning, LLMs, such as Grok’s image generator, which produced the wrong kind of images. Everyone objected to the videos Grok made. Unsafe AI generally leaps from LLMs and excessive training. Another unsafe AI is possibly Mythos, developed by Anthropic itself, which, its own CEO says, can pose a threat to cybersecurity. Other unsafe AIs are agentic AI bots that can find ways through code to machines. Another kind of unsafe AI is one that generates its own language. Machines talking to machines in unrecognizable symbols is another example of unsafe AI.
With this much compute time provided to Anthropic. The maker Dario Amodei must comply with the standards of safe AI for mankind. Projects such as Mythos must be developed within constraints. And other future projects that could harm mankind must not be sent for testing and should be discarded.
Unsafe AI can harm the growth of innocent children. Children must hence be not confronted with AI of unsafe kind. The classification of safe and unsafe AI is broad and can be made using a mature classification framework.
We are in a race to build human superintelligence first, but we have forgotten that such a superintelligence can harm human lives in ways not comprehensible in current times; leave aside children, even adults’ lives can be hampered by superintelligence.
So why do we need AI to make a superintelligence?
Why are people still testing newer models for superintelligence?
It’s become a weird race now, where all AI companies are trying to make new models for AGI and superintelligence, with new features, all to gain new users?
New users use the new models and ask, “Why are we using it?”
Artificial superintelligence is a threat to mankind.
However, safe AI, such as robotics, can help mankind, say, build life on another planet. Robotics with no access to superintelligence AI is safe AI as long as robot follows its protocols.
Other safe AI can be bots or robots working remotely on another planet.
Other safe AI includes AI robots used in human-operated remote medical surgeries.
Hence, not all AI is bad.
So, we must understand the difference between safe AI and unsafe AI.
Not all AI is bad. An AI that runs standalone on your phone and can classify your crops as rotten or good is safe.
While an AI that can hack your systems via agents and other new AI is an unsafe AI. Mythos, for example, as Anthropic itself has noted, was unsafe and is therefore restricted in use by Anthropic itself.
So, do understand what is safe AI and what is unsafe AI.
Now, Musk had a bad time with Grok and Anthropic had a bad time with Mythos, so what can be bad time for their partnership. Let’s hope they build safe AI, for enhancement of mankind either on Earth or outside Earth.
Thank you for reading.
Subscribe for updates.
Artificial Intelligence, Computer Science & Mathematics 