AI is a hub of discreet technologies, development in its subset fields has been resulting in the coolest milestone achievements of AI. AI might not be as smart as humans but it’s already proved that it can beat humans at some point of purpose. Sometimes it wowed us with its exceptional performance and sometimes raised so many worries with its unlimited capabilities.
Development in AI has created new hopes for the future and a lot of worries at the same time. Questions like ‘are we really need AI’, ‘does AI take away jobs’, and ‘is AI good or bad’ are a few that repeatedly striking ethicists, and it’s completely a different topic. If you are keen to learn more about AI, refer to our previous AI insider article which I made it in a clear and simplistic way about ‘what’s and how’s’ about AI.
Milestone achievements of AI:
Though, AI is rapidly developing and making its mark on every industry. Here are the coolest achievements of AI for the decade.
If you’d ask me “what is the most notable achievement of AI for the decade?”
I’d say “that its accessibility, AI caught much attention and got enough pace in this decade. Once it was a science fiction story and now it’s a reality that took part in our everyday life. And, interestingly most of us don’t even know it.”
In the most common way, we used to use Google every day. ‘What you’re searching and what you’re interested in’, Google knows much about you than that you think what Google knows about you. Not only Google but there’re also a lot of internet companies out there that already incorporated AI into their businesses.
And, these are smart enough to predict ‘what are the things you most likely to buy online and what are the news and topics that you’re much interested in’, this data will help marketers to target their customers/audiences at higher conversion rates. AI allows social media to show the right feeds/posts from the right people to reach targeted audiences. Also, what music you’d like to hear, the best fastest route while commuting and voice assistants are some of the examples of AI we use in our routine.
2. AI achieved human-like thinking capabilities
Thanks to deep learning.
Deep learning is a technological approach to build an artificial brain that inspired by the structure of a human brain. In deep learning, artificial neural networks (ANN) simulate the process of biological brain cells (neurons) which will enable human-like thinking capabilities to a machine.
A group of researchers from Google and Stanford University described how they build a deep learning model – AlphaZero, a breakthrough achievement towards building an artificial brain that masters chess and shogi games through self-play in a research paper. They claimed that “AlphaZero is based on general reinforcement learning algorithm and can be trained and learn itself by self-play.” Also, they claimed, “AlphaZero acquired ages of human knowledge in chess, shogi and go games in just a few hours.” This is the most significant achievement of AI that strengthens our hopes towards achieving artificial general intelligence (AGI) in the near future.
In contrast to machine learning, deep learning techniques skip manual data labeling, a process to prepare data sets for training machine learning models, which is very expensive and time-consuming. Deep learning algorithms can learn on their own by automatically analyze and label the data by its various features and properties. Which in turn, result in accelerated development in the field of AI and opens up a new world of possibilities. Such as self-driving cars, image caption generation, visual manipulation, voice assistants, and machine translation are a few of which we’ve witnessed over the past decade.
3. AI can create and train AI
AI has achieved much significant success this decade and all of them are relies on human-expertise. Though, an automated machine learning model basically developed by human-experts, is capable of creating their own child AI itself.
Yes, AI can create AI without human involvement. A group of researchers at Google claimed that they build a Machine learning model called AutoML based on deep neural networks that can create child AI, technically Machine learning models called NASNets which outperformed all the other man-made AI in the world.
NASNets are 85% more accurate than any other AI built by humans in recognizing objects in the images, which could be another breakthrough in computer vision. Also, adaptable to other disciplines of AI. As AutoML automates the process from initial data pre-processing to deployable machine learning models to real-world problems, it saves the much time and cost in building AI solutions for real problems.
4. Self-driving cars
A fully autonomous car is one of the serious and major agenda for the AI over this decade. More companies are engaged in testing and developing to bring driverless cars to the traffic-clogged streets. From Tesla’s Autopilot to Waymo and Pony.ai’s commercial self-driving taxi services are a few of the major milestones in development in self-driving cars.
The whole automobile industry looking towards deep learning – a significant technology of AI – in order to replace human drivers. Experts believe in “Motors with applied AI and discreet technologies of its subset fields will result in more breakthrough innovations and paves the path to build a fully autonomous car.” Deep learning will help vehicles in perceiving the environment such as voice-recognition, motion-detection, traffic signal recognition and predict the behavior of pedestrians and other vehicles on the road.
Though, a fully autonomous vehicle is not yet achieved that truly means as it called. Not only the automobile companies but also the tech companies from Silicon Valley are trying hard to build brains for vehicles. And, you’ll notice it if you closely following to the tech updates, Google, drive ai, Yandex, and Nvidia are the few.
When can we expect a fully self-driving car to be a part of our daily lives?
There is no such particular date and time defined, but…
Some of the government’s initiatives, Increased fundraising, and exclusive laws for driverless cars are the major signs that make it clear that “there is no need to long await.”
5. AI Defeated Humans
All that AI is available today is based on narrow AI which can be built and trained after to perform a specific task only. Though what? AI achieved a superhuman level in its specified task and became strong enough to out-perform humans. From IBM’s Deepblue in Chess game, Watson in ‘jeopardy’ game show, Deepmind’s AlphaGo in Go game to recent Facebook’s Pluribus in a multiplayer poker game, AI proved itself many times throughout the decade.
6. Voice Assistants and chatbots are at your service 24/7
This is the perfect example of AI accessibility. We all use voice assistants or have tried for at least once. If you’re a regular user then you’ll most likely to closely notice the evolution of voice assistants. Here are the few:
- Deeper integration into the operating system and third-party apps will allow you to access the settings, notification, and services of OS & third-party apps.
- Voice assistants have learned to read aloud your emails, messages, and news updates for you.
- Gain the freedom to choose a second language option to assist you in your native language.
- More personalization for better assistance – as they trying to know all about you. ‘Hay Google – let’s go to work’ then it will show the direction to your office. Or just say ‘call mom’ then it will dial your mom’s number. Say ‘tell me about my day?’ or just ‘what’s on my calendar today?’ to know your schedule.
- Streamline conversations by skipping wake words for better conversational and natural experiences.
- Voice recognition for better security.
And, chatbots became the mainstream customer support for business to enhance their user-experience. Facebook introduced an automated answering messenger chatbot for business accounts. Almost every business and banking website you visit will show a chat box with “hello, how can I help you?’ message.
As we discussed in our earlier post of AI Insider, the real success of AI is achieved when AI algorithms become more efficient (speed and accuracy) and capable of performing a general range of coordinative tasks such as conversation, movement, planning, and problem-solving just like the humans. Industry experts described automated machine learning (AutoML) is a fast track to developing general-purpose AI (AGI). But, some have expressed their worries about ‘An I building AI without human involvement’. Yet there is so much see, and the excitement and the fear continues…