Pascal's Wager

-


 I haven't written in quite some time, and I thought about pushing it even further. So be it, let us push further into the future. Here is what I wrote for a newly-learned topic. 
_________________________________________________________________________

Capabilities: 

Algorithmic and compute advancement 


Social: 

Socioeconomic changes 


Systemic: 

Policies, regulations, and incentives. 

_________________________________________________________________________


Question: What will AI systems look like in 10 years and how will current AI systems develop to get there?


AI Systems Landscape by 2032 — evaluating current and future trajectories 


Before I begin, I would like to address my critical shortcomings in my prediction. First, I suspect that I will be taking a convenient sampling, basing my predictions mainly on artificial intelligence precedents and their respective characteristics. Second, I will approximate this landscape with a limited number of variables. There might be variables that we have not even heard of, such as sudden changes in the political climate regarding AI developments in the future. This report acts as an idea/opinion bank. 


There are a couple of sub-questions that I want to answer in my report: 

  1. What are the current AI types, usages, and advancements? 

  2. How do I define transformative AI? 

  3. How will transparency/open sourcing affect the development of AI systems? 

  4. What is an example of how I would predict [insert scenario]?

  5. Is an intelligence explosion achievable in the next ten years? 


Current AI types, usages, and advancements


Currently, narrow AI has been progressively able to surpass human capacity— either by obtaining larger memory storage or higher information retrieval rate at an instance of time. For example, Open AI currently uses supervised learning, meta-learning, and generative models. However, there has been a report on how the recent successor of GPT-2, GPT-3, could be approaching the capability limitations of predictive language models (Sobieszek et. al 2022). 


If GPT 2 was released in early 2019, and GPT 3 was released in the middle of the consequent year— then within two years, or by 2024, it is foreseeable that predictive language models will have reached their threshold cognitive performance. However, it is only one type of narrow AI. 

Using a combination of systems neuroscience and machine learning, DeepMind’s reinforcement learning algorithms culminated in a sensational headline in 2016 (Silver et. al 2016). AlphaGo, an algorithm to solve Go, has since been refined to MuZero in late 2020. MuZero is a general-purpose algorithm that masters Go, chess, shogi and Atari without parameters— it did not need to be told the rules. Planning winning strategies in unknown environments highlights DeepMind’s and other initiatives’ algorithms to implement an adaptable AI upon narrow AI. It is broadening the focal point. 


From 2016 to 2020, MuZero has sinced improved drastically. From this example, I predict that games with exhaustive steps and walks would be able to reach its threshold by 2024. 


While we can continuously improve on narrow AI, the impending topic is how and when a generalist AI agent will come about. 


Defining Transformative AI 


Transformative AI is a type of AI that will radically change the way society operates to the common individual, on a daily basis. Honing in on the topic of widespread AI use, from the graph below, we can see that in 2019 Capterra Top Technology Trends Survey yields a 25% US small businesses intending to use AI and ML in the next one to two years while 27% plans to evaluate some more. 


Figure 1. AI & ML usage 



We can thereby infer a great majority of the small businesses, and can deduct that large industries just by necessity of AI and ML support, will use AI and ML when given the opportunity. Transformative AI, then, would not need too much time to be incorporated once AGI appears. 


[ I didn’t have enough time to find context for AGI, will continue tomorrow ]  


The impact of increasing transparency to AI systems & how it will affect our timeline


The feasibility of increasing transparency to AI systems is high, and even higher in the next ten years. Deepmind’s AlphaFold is an example of increasing transparency to the general public. AlphaFold, an algorithm trained from a public repository of proteins, aims to predict protein structure. Similarly, Google Brain’s TensorFlow was also open-sourced to whoever like to access the data. 


This will affect our timeline by shortening our AI systems growth, by disseminating more information to the public. However, we must also implement more appropriate publishing norms to avoid information hazards from open publication. The most risk comes from what Bostrom would define as data hazards and idea hazards, which multiple rounds of editing and converting to layman terms would fix (Bostrom 2011). 



Is an intelligence explosion achievable in the next ten years? 


Based on my understanding of an intelligence explosion, it will happen as fast as the next recursively intelligent-refining machine is to be invented. When this machine will be invented is a debate. Now, Paul Christiano says that economic output can be used to model technological acceleration. He is measuring it via a logarithmic transformation of GDP over time. Inferring from his model, technological acceleration will just keep on increasing. However, this hyperbolic growth is contingent on continuous, unrestricted limiting factors. Unless a transformative AI can be developed, the leap-frogging growth simply would not be true. Therefore, an intelligence explosion, I hypothesize, is not achievable in the next ten years. 


Figure 2. From Paul Christiano


Comments

Post a Comment

Popular posts from this blog

ArtificiaI General Intelligence Safety Fundamentals Notes

-
Image
AGISF - Week 1 - Artificial General Intelligence Reinforcement Learning Foundation models : models trained on broad data that can be adapted to a wide range of downstream tasks. Much of foundation models are enabled by transfer learning and scale. In which scale required:  1. GPU throughput and memory capacity, both increased 10x over the last four years.  2. Transformer architecture  3. Availability of training data  Homogenization: all the defects of the foundation model are inherited by all the adapted models downstream.  Four Background Claims by MIRI, 2015 CE  This post aims to illustrate MIRI's mission objective.  1. Interactions between disparate modules is the general intelligence that we can not replicate in code, yet. MIRI suggests to take heed.  2. Given that the Church-Turing thesis hold, then computers can replicate the functional input/output behavior of brains as physical systems--- they might supercede soon.  3. Intellectual advantage will create incentive to
Read >>

A year, introspectively summarized

-
Image
               A year, introspectively summarized      A primal and gravitational feeling is the feeling of belonging--- though the whole sense of "purpose" is artificial, defined deliberately by people who want others to emulate them. Emulation stunts creativity, just as the proclivity to belong stunts agency. I am delighted by the platform that I am given to practice my agentic self, now and continuously onward.       There will be a time when t he basis of your understanding, you learn, was fabricated---even some of your strongest beliefs need reinterpretation, the very framework under which you operate proves inflexible and restrictive. Perhaps the best remediation strategy then is desensitization of the action-oriented part of your mind. You must compromise scaling for precision.       In January, I wrangled with partial differential equations and AI Safety fundamentals course with BlueDot. In February, I wrapped up two of my internships. In March, I explored much of Los
Read >>

Vicissitude

-
Image
Vicissitude Did the heavens hang higher to aggrandize your vicissitude? From the primal vines you serpentine ushering empires and visions sublime   If thunder be your heartbeat and determination then let lightning linger in your mind from deprivation unbounded by ecclesiastical measures, you drape your golden body over boundless wonders  The dreamers drown in idolization and ideals but sharper is the blade of guilt the thirst for ravishing power and the departure from righteousness your celestial eyes close and halt On the descent, since the fall the leathery wings enshroud your undying faith In your shadow woke Wicked Mephisto  down is the plunder into ecstacy  but who can share your burden of villainy?
Read >>