2023 ?? wrap-up

Its amazing to see how AI adoption happened in the year 2023. With all the Generative AI use cases and new products, last year felt like evolving at an unprecedented speed. Reading through all the stories and posts by AI Influencers, Entrepreneurs, Domain Experts, and literally "Common People", was so overwhelming. I would call the year 2023, an year of defining the new AI tech-disruption of this decade.

The world has seen a lot of things, from AI taking a center stage of all the drama and limelight to a rollercoaster year for funding in the AI sector. Weather it is a whopping $10 billion investment from 微软 to OpenAI , or the race of jumping in or enhancing GenAI by tech giant's like, 谷歌 , 英伟达 , Google DeepMind , Meta , IBM , Snowflake , Databricks , Anthropic , Notion , and 100s more.

Products like OpenAI 's ChatGPT, Midjourney , 苹果 Vision Pro, Adobe Firefly, and IBM Quantum System Two, has shown the world a new possibility of technology + innovation access. And at the same time the larger risks in questions like DeepFakes, AI-powered drone swarms, 亚马逊 's Rekognition software, Automation anxieties, AI Job Displacement, and many more, have raised the concerns over responsible + ethical AI use. And all of this to digest and picture for the year 2023 is hell lot of a task. But let me try break it down in the best form I can, and by starting the biggest buzz of AI funding's.

Table of Contents

1. Top AI companies that raised the biggest funding in 2023
2. The major advancements in the AI technology use cases
3. AI Controversies in 2023
4. AI things happening under the hood — the tech-stuff
5. A dynamic year for M&A in the AI space
6. My favorite ?? LinkedIn posts from 2023
7. My Favorite ???? GenAI Reports and reads of 2023
8. The Closing Note

1. Top AI companies that raised the biggest funding in 2023

The major funding's in AI happened around the Generative AI this year, and that's an uncontested fact. If you are a LLM based company, you would have raised something for sure, it's that kind of an year. Let's look at some of the most watched funding:

1. OpenAI ($10B) — @Microsoft confirmed a multiyear, multibillion-dollar investment in OpenAI, the maker of ChatGPT and DALL-E, in late January. The exact amount is undisclosed, but it's speculated to be up to $10 billion.
2. Anthropic ($7B) — Anthropic, a competitor of ChatGPT, raised nearly $7 billion. 亚马逊 invested up to $4 billion in September, giving them a minority stake. This included an initial $1.25 billion, with a possible extra $2.75 billion. Anthropic will use Amazon Web Services (AWS) for its AI models. Also, 谷歌 invested up to $2 billion in Anthropic in October, starting with $500 million, and potentially adding $1.5 billion more. This followed Google's earlier investment of $300-400 million in the startup.
3. Inflection AI ($1.3B) — Inflection AI raised $1.3 billion for AI development. Valued at $4 billion, it is creating a large AI cluster and an AI assistant named Pi. The funding round was led by Microsoft, Reid Hoffman , Bill Gates , Eric Schmidt , and 英伟达 . Inflection, founded last year (2022) only and has now raised over $1.5 billion. Mustafa Suleyman , the co-founder was Google DeepMind CEO.
4. Metropolis Technologies ($1.7B) — It's a vehicle parking startup using computer vision and AI, raised $1.7 billion in debt and equity. Funding came from Eldridge Industries and 3L Capital, included $1.05 billion from a Series C offering and $650 million in debt financing. The funds were primarily used for the acquisition of logistics firm SP+ (SP Plus) in a deal worth about $1.5 billion, marking the year's largest M&A transaction by a VC-backed company. This deal surpassed Databricks' $1.3 billion acquisition of Databricks Mosaic Research , a language models training startup.
5. Databricks ($685 million) — This data analytics firm with AI enhancements, raised over $685 million in a Series I round led by T. Rowe Price Associates. This funding valued the company at $43 billion, up from $38 billion in 2021 following a $1.6 billion Series H round. 英伟达 , increasingly investing in AI startups, also participated in this round. Databricks recently reported over $1.5 billion in revenue with more than 50% year-to-year growth.

The were many more companies who raised funding this year, and in the AI world 2023 established the strong foothold AI is going to embark. Beyond these few of the other funding's that we can't miss were:

• Abound ($601M) - The open banking start-up played the AI card and raised money, Aleph Alpha ($500M), SandboxAQ ($500M), Mistral AI ($487M), Adept AI ($350M), Cohere ($270M), Builder.ai ($250M), Hugging Face ($235M), Genesis Therapeutics ($200M), Asimov ($175M), and god knows how many more...

Also, in this funding spree, some of the companies that I have been following and reading around were Midjourney , Tempus AI , Dataiku , Highspot , AlphaSense , Shield AI , Grammarly , Eightfold , Moveworks , Abnormal Security , Runway , Labelbox , Interactions LLC , Synthesia , Character.AI , AILY LABS , Captions , CentML , Durable , Entos Pharmaceuticals , GPTZero , LangChain , PolyAI , WellSaid , SaaS Labs , among many others.

2. The major advancements in the AI technology use cases

2023 has been a year of breakthroughs across various AI fields. Ideally, AI has taken a center stage this year. Here's a glimpse into some of the most impactful use cases:

• GenAI Image Generators — I couldn't start with anything else, tools like Imagen from 谷歌 AI, Midjourney , and OpenAI DALL-E, continue to push the boundaries of image generation, creating stunningly realistic visuals from text prompts.
• Text-to-Video — Startups like Runway and Pika have emerged, allowing users to convert text descriptions into moving images. This democratizes video creation and has the potential to revolutionize content creation.
• Code Generation — OpenAI 's Codex and GitHub 's Copilot assist developers by predicting code and automating repetitive tasks. This improves programmer productivity and accelerates software development.
• Personalized Education — AI tutors like CENTURY from 美国卡内基梅隆大学 personalize learning experiences for individual students, adapting to their pace and needs. This promises to improve educational outcomes for everyone.
• Healthcare Revolution — Deep 6 AI and other companies utilize AI to analyze medical data, leading to faster diagnoses, personalized treatment plans, and drug discovery. Specially with this year, AI is literally transforming healthcare at the clinical level.
• Climate Change Solutions — This is very special to me, as I have been doing lot of research and modeling in this field, and would dedicate a newsletter to this topic for sure. AI models have advanced more towards predicting weather patterns, optimizing energy consumption, and even designing sustainable materials. Today, the fight against climate change is increasingly powered by AI-driven initiatives.
• On-Device AI — The likes of Amazon Web Services (AWS) , Google Cloud , Microsoft Azure , etc. have extended AI models to enable Edge computing for processing AI directly on devices. This enables real-time features like translation, image editing, and anomaly detection, even offline.
• Natural Language Processing (NLP) — LaMDA from 谷歌 AI and Megatron-Turing NLG from 英伟达 shows the outstanding progress in language understanding and generation. And this has the potential for way more natural human-computer interactions.

While these are just some key highlights, 2023 has seen some major progress in robotics, autonomous systems, and other AI subfields. As research is progressing and partnerships between academia and industry is strengthening, the baseline to reshape our world using AI is established.

3. AI Controversies in 2023

Whenever there is a disruption, there are flipsides that are also important to consider. GenAI has already acted as a double-edged sword because of Deepfake proliferation, Bias and discrimination, and Copyright and ownership issues. Beyond that the fears of Job displacement, Algorithmic bias in hiring and recruitment, and the ethics of AI surveillance in workplaces, have grabbed most of the eyeballs.

There are peculiar situations and their follow-up questions that people have faced and asked, like weather to use AI for predicting crimes or to identify potential criminals based on data analysis. One of the most famous one, when on the April 11, 2023, in Pakistan, a judge in a session court used OpenAI 's ChatGPT to help decide whether to grant bail to a 13-year-old accused in a case. The court mentioned using ChatGPT's assistance in its final decision:

Can a juvenile suspect in Pakistan, who is 13 years old, be granted bail after arrest?

The OpenAI 's ChatGPT replied:

Under the Juvenile Justice System Act 2018, according to section 12, the court can grant bail on certain conditions. However, it is up to the court to decide whether or not a 13-year-old suspect will be granted bail after arrest.

But, beyond Generative AI, AI in general have also seen a lot of tensions, like the Black box problem, wherein many AI algorithms are still opaque and difficult to understand. And this makes is hard to identify and address the bias, errors, and potential human harms it may cause because of its decisions.

On a similar side, the competition of dominating the AI world between tech-superpowers have been like a fast-and-furious. Everyone knows how 微软 , 谷歌 , and Meta are fighting in this race. And the angles of trade war between USA, China and other countries raises concerns about a potential AI arms race that can increase international tensions further.

One of the major discussions over public forums have also been about the access and control of AI technologies. Unequal access to and control over AI technology can worsen existing economic and technological disparities between nations. Australian MP Julian Hill raised concerns in the national parliament about the potential negative impacts of AI's rapid development. In his speech, partially authored by AI itself, he highlighted risks like widespread cheating, job losses, unfair discrimination, spreading misinformation, and the danger of AI being used in military technology beyond our control.

ChatGPT is scary good. We are not far from dangerously strong AI — Elon Musk

In March 2023, over 20,000 signatories, including tech leaders like Yoshua Bengio , Elon Musk and Steve Wozniak , called for a halt to large AI projects due to societal risks. Geoffrey Hinton, a key AI figure, left 谷歌 in May 2023 over concerns of AI surpassing human intelligence. That same month, a statement from AI experts highlighted the need to address AI's potential and existential risks.

4. AI things happening under the hood — the tech-stuff

?? This part is bit traversal if you aren't an AI specialist. For AI power users, I would recommend to skip this part and jump on next section directly.

At the core of AI advancements there were swift improvements and the baseline of overall AI technology has setup a new level of benchmarking. Starting with the techniques like gradient checkpointing and sparsity has allowed training on larger datasets with limited resources. This has lead to the rise of more powerful and balanced LLMs like OpenAI GPT-4, 谷歌 's PaLM, and 英伟达 's Megatron-Turing NLG.

Also with the integration of text, audio, and visual data into LLMs, we have seen a deeper understanding of the real world enabling multimodal communication capabilities. Good example of these are OpenAI 's CLIP and 谷歌 's Imagen.

Especially, due to the public access to interact with these model's, the platforms have earned enormous amount of data interactions which had helped them to further improve their core model. A fair example of this is how ChatGPT code solving abilities changed with time, the developers asked questions, and if they get follow-up errors, they posted it to ChatGPT, which further did this activity of problem solving.

Some of the main highlights if I have to pick it would be these:

1. Integrating symbolic AI with deep learning techniques have made the systems to understand the casual relationships, which has helped models to make more logical interfaces. This was specifically demonstrated with the Google DeepMind 's AlphaFold for protein structure prediction.
2. Again the integration of symbolic AI with statistical methods. The IBM 's Project Debater was made capable of engaging in argumentative dialogue using this method. And this has shown the ability of flex reasoning and knowledge representation.
3. Recalling from the Google DeepMind 's Gato, which has learned a variety of skills ranging from playing Atari games to controlling robots, all with minimal interaction with its environment. And it was possible because of the reinforcement learning techniques like sample efficiency and meta-learning, that helped AI to quickly learn complex things using lesser interactions.
4. Next mind blowing thing was the hardware architectures which reflected the human like brain structure. 英特尔 's Loihi was a demo out of this world promising more energy-efficient and parallel processing for AI tasks.
5. Areas like Quantum computing specifically for AI, and the Edge AI with decentralized processing opened up the doors in the domain of materials science and drug discovery. This decade would see a next level healthcare use cases of AI that we can't imagine today.
6. Methods like attention mechanisms and feature attributions have really helped us to understand on how AI models arrive at their decisions, and this is going to be a key for improving trust and transparency of AI's future.
7. Techniques like counterfactual analysis and debiasing methods have now gained good popularity and is being used by developers more rigorously to address potential biases in training data and model outputs.
8. Frameworks like TensorFlow and PyTorch , along with Hugging Face have given power to developers and researchers to accelerate innovation through easy entry.
9. Cloud providers like 谷歌 , Amazon Web Services (AWS) , and Microsoft Azure offer readily available AI tools and infrastructure, which makes AI access easier for businesses and individuals without having exceptional technical expertise.
10. The 谷歌 's AdaFactor optimizer to address the issues of vanishing gradients and exploding learning rates , so that there is an improvement in training stability and convergence for LLMs and other deep learning models.
11. Google DeepMind 's Stochastic weight averaging (SWA) technique which improved the generalization and also helped in reducing the model variance by averaging parameters over training iterations. And this has lead to more robust performance on unseen data.
12. Techniques like model-agnostic meta-learning (MAML) enabled the AI agents to learn complex tasks with just a handful of examples. And due to this rapid adaption to new scenarios has opened up.
13. Multi-agent reinforcement learning (MARL) coordination algorithms, like DeepMind's CommNet, has enabled multiple AI agents to cooperate and tackle complex challenges in environments with shared goals and potential for competition.
14. Dreamer from DeepMind allow learning from large offline datasets and seamlessly adapting to real-world settings, unlocking the potential of simulation-based RL for robotics and other embodied AI applications.
15. Facebook AI's contrastive approach, which offered explanations that highlights why a model chose a particular output compared to alternative possibilities. And this provides a richer insights into the decision-making process.
16. Lastly, not to miss, the methods like IBM's Fairway framework to assess potential biases in models by analyzing how their outputs would change under hypothetical counterfactual scenarios.

To be fair, these are among the handful stuff that kept me curious and pushed me to explore, but there were quite few more that have been utilized more maturely and more counter-intuitively in 2023.

5. A dynamic year for M&A in the AI space

With several notable deals signaling consolidation and shifting priorities within the industry. Here's a roundup of some of the most important acquisitions as per me in the AI sphere that took place this year:

1. Neeva (February 2023): Snowflake , a cloud data management company, acquired Neeva, a search startup founded by two former Google employees, for approximately $150 million. The acquisition will help Snowflake's enterprise search capabilities through Neeva's generative AI expertise.
2. Myst AI (acquired by Snowflake) (January 2023): Another Snowflake acquisition. Myst a company developing an AI platform for time-series data analysis in the energy sector. This will strengthen Snowflake's offerings in AI-powered business forecasting.
3. nod.ai (October 2023): AMD acquired nod.ai , an open-source AI software company, to enhance AMD's AI software capabilities and to optimize AI models for its hardware products, primarily graphic cards.
4. GamePlanner.AI (August 2023): Airbnb bought GamePlanner.AI , a travel planning startup utilizing AI for personalized activity recommendations. This acquisition will help Airbnb's to improve their travel experience offerings and will also expand its reach in the activities and experiences sector.
5. Falkonry (August 2023): 艾菲诗软件 , an enterprise software company, acquired Falkonry, a developer of AI-based time-series data analytics for asset management. With this acquisition Falkonry's technology will be integrated into IFS's ERP platform, which will boost its enterprise asset management services.
6. LeapYear (acquired by Snowflake) (February 2023): Snowflake acquired LeapYear, adding its differential privacy technology to its offerings. This technology helps protect user data while enabling data analysis.
7. Modin (March 2023): Palantir Technologies , a data analytics company, acquired Ponder, a NLP startup, to enhance its data exploration and analysis capabilities.
8. Fig (June 2023): 谷歌 acquired Fig, a collaborative writing platform, to enhance its Google Docs features and potentially explore integrations with its AI technologies.

Overall, if you see, the 2023 AI M&A landscape shows that the industry is maturing with continued investments and strategic acquisitions which are aimed at strengthening specific capabilities, enhancing user experiences, and exploring new avenues for AI application across various sectors.

6. My favorite ?? LinkedIn posts from 2023

Here are some of my handpicked favorite posts on LinkedIn. Did I miss something? Please comment and let me know ??

7. My Favorite ???? GenAI Reports and reads of 2023

1. The AIdea of India by 安永
2. Impact, Opportunity and challenges of Generative AI by INDIAai
3. A new era of generative AI for everyone by 埃森哲
4. Generative AI is all the rage by 德勤
5. The state of AI in 2023: Generative AI’s breakout year by 麦肯锡
6. Finding value in generative AI for financial services by MIT Technology Review
7. Generative AI: HYPE, OR TRULY Transformative? by 高盛
8. Generative AI Surveys by Gartner
9. Generative AI: From buzz to business value by KPMG
10. Harnessing the value of Generative AI by 凯捷咨询
11. The Future of Generative AI: An Analysis of the Leaders, Opportunities, and Threats by AlphaSense
12. Australia’s Generative AI opportunity by 微软
13. Generative AI - The rise of the machines by 汇丰
14. The Crowdless Future? How Generative AI Is Shaping the Future of Human Crowdsourcing by 美国哈佛商学院
15. Generative AI in the manufacturing Industry by 高知特 Cognizant

8. My Closing Note

This year have been a fantastic year for the AI landscape. And 2024 is going to put this all in use for building advanced use cases. The year 2023 will be remembered as a time when AI ceased to be a distant, futuristic concept and became an integral part of the present. With these massive investments in AI startups to all the controversies of generative AI, NLP and LLMs have made itself a dominant AI space the whole year.

Looking ahead, the possibilities are endless. As AI will further evolve, we can expect even more transformative changes across various sectors. The key would be to navigate this in a balanced manner keeping the innovation growing, but alongside being mindful of the ethical and societal implications it has.

So, 2023 has set the stage for a new era in AI, one that promises to be as exciting as it is unpredictable. And as we have stepped into 2024, let's carry forward these learned lessons and continue to explore the amazing potentials of AI, but with a commitment to create a future that benefits all of humanity.

????????, stay tuned and subscribe and follow to get notified

? Read A comprehensive list of 2024 AI predictions ??

?? Thank you for reading this. auto9mous is my personal bi-weekly newsletter wherein I explore and dive deeper into the latest advancements in digital transformation, new tech, data science, AIML, digital twins, AR & VR, and more.

My newsletter is your backstage pass to the coolest and most cutting-edge tech trends out there. From mind-blowing innovations to inspiring success stories. Join my auto9mous newsletter by subscribing it, and keep yourself updated in this ever-evolving world of technology.

?? Blog: https://abhilashshukla.com

?? X (Formerly Twitter ): https://twitter.com/AbhilashShuklaa

?? Want to connect? https://calendly.com/abhilashshuklaa/catch-up

?? Book Summaries, and Case Studies: https://abhilashshukla.com/books/